If you missed it, you can read:

• #1 - Financing Risk : a brief history of Venture Capital

• #2 - Chance, the Wrapper: VC as a Financial Product & Asset Class

• #3 - Sum of the Part(ner)s: Why VC Partnerships Fail

• #4 - Backing Great Founders : The Biggest Cliché in VC

• #5 - The Craft: How VC Firms build edge

• #6 - Runway, Revenue, Regime: VC & the State

“The world has changed. I feel it in the water”

The Lord of the Rings: The Fellowship of the Ring (2001)

The same eerie feeling that Galadriel evokes has started to seize me these past months.
And if you’ve been following this series, you should also have sensed some tension building in Venture Capital throughout the previous editions. The change is afoot.

What do we call “venture capital”, really?

We’ve gone to great lengths to describe it as an endeavour of financing risk, as an asset class, as a bunch of more-or-less successful firms, as a mission to back great founders, as a craft or as a state-sponsored innovation mechanism. But maybe another way - the simplest - to look at it is to take the most ordinary unit of this business (deals) and to define what venture is through it.

So let us have a look at what has been called a “venture” deal in the past year:

• Anthropic raised $65bn on a valuation approaching a trillion dollars post-money from a syndicate roster that mixed Sequoia, Greenoaks, Altimeter and Coatue with Blackstone, Brookfield, public managers and hyperscalers. The money is to buy compute, advance its research and secure its market position. (0)

• Ineffable Intelligence, founded in late 2025 by former DeepMind researcher David Silver, raised $1.1bn pre-product thanks to a (jam-packed) resume and a moonshot pitch.

• Arga Labs (YCP26), a (random) company building a solution “to test and train AI agents against realistic, stateful twins of the external systems they interact with - without ever touching production” has raised $10m from General Catalyst, likely to buy distribution and headcount. This is your “typical” 2026 seed-stage round (1).

• Somewhere, an upcycled e-bikes marketplace is trying to scrape together $2m for eighteen more months of survival. Companies with real revenue, teams and operations are failing to raise bridge rounds. A biotech is raising nine figures in order to reach a clinical milestone that is four years away. We don’t need precise examples for those ones: you know them.

• And at the other end of the spectrum, there’s Medvi, a two-person GLP-1 telehealth business that did $401m of sales in 2025, is tracking toward $1.8bn. Medvi isn’t a deal because it never raised. In shape, it looks like what used to be venture-backable and venture-scale but in fact, it no longer needs the financial product.

These transactions are all “venture deals” (s/o Brad Feld) in their own right: they share the same contractual format with preferred stock or a SAFE, a cap-table, a stock-options pool, a board. But in no way do they share risk, duration, path to cash, return profile or realistic exit. We’re somehow grouping very different transactions together under the same “venture” flag - “venture” alternatively pointing at risk, stage, technicality or innovation.

Using the word to mean different things is fine. The problem lies in what we keep deducing from the word: that the single label describes a single financial product, when clearly it doesn’t - or shouldn’t.

Indeed, there is a gear shift in venture happening in several layers at once: the deals have diffracted (2), which is the observation above. But the wrappers have diffracted too: deal clubs, syndicates, deal-by-deal SPVs, solo GPs, hyperfocused verticals, corporates and family offices going direct, megafunds at one end and micro-funds at the other, yaddi, yaddi, yadda …

Nothing new under the sun: both observations are, by now, well made in VC literature (Kyle Harrison for instance has made it here, here, here and here). What’s interesting and rarely noted is how both movements coordinate: new deal and new vehicle-types are just the same mechanism observed at two ends of the chain.

Let’s take vehicles: in Edition #2, we gave an equation for the typical venture fund: Fund = Individuals × Access × Strategy × (1 + Past Performance)^(Iterations).

That equation worked when the deal shape was canonical enough to make sense as a bundle. Indeed, as we’ve previously argued, nothing about the 10-year fund makes it a law of nature: it happened to be the best-fitted solution to the standard VC deal shape. You probably know that shape intimately: pour a few million into a company, have entry prices that increase predictably across stages, and exit in 5 to 8 years with re-ups along the way (3). That well-known shape demanded all factors of the equation at once (Individual, Access, Strategy) in roughly comparable measure. Which is why they could be bundled, priced at 2/20 and sold as a decade-long subscription.

The thing that’s novel in 2026 is that the deal shape no longer exists: the width of the spectrum grew several orders of magnitude. Out of one historic deal shape now came many (hence the title). And alongside it, it has become possible (and cheap) to stop selling the unused parts of the fund equation.

When a transaction still needs every factor at once, the fund remains the right object. When it needs one or two, a firm that carries all of them is bundling work that does not get used, a useless part of the value proposition. And when that happens, someone somewhere will sell fewer of those factors, more cheaply, to the same clients - in our case, the LP.

The same unbundling that happened to many industries is now quartering Venture Capital.

And so what this edition explores is that venture capital is like white light: there are several wavelengths travelling together and called one colour because nobody had yet held up a prism. But let’s put that light through the prism, shall we?

The New Deal(s)

Deals used to be described with recognizable terms that immediately came with a mental map and handy frameworks. “Seed”, “Series B”, “SaaS”, “deep tech” meant something, and any practitioner could infer the rest: price estimate, incurred risk, check size, current headcount, exit options.

However, the mental map is now wrong. As it was some day discovered that California was not an island and the maps were wrong, the two main legends of the old map - stage and round size - no longer serve their purpose (4). We’ll argue that the new legend that actually works is closer to the actual operations of the company: cash burn.

On stage

Arguably, a deal stage never really conveyed information about size. You could raise 10x more at the same stage: YouTube raised a $3.5m Series A in 2005, Amazon raised an $8m one in 1995 and Google a $25m one in 1999, way before any VC in the Valley had heard about neural networks. But it did convey some information.

In Edition #4, we drew a line in the sand: at pre-seed and seed, the founder is the asset because there is nothing else to underwrite. By Series B, the asset is now the company: it has KPIs, a team, (usually) a product that met its market, etc. Everything in between is a different shade on the spectrum - we’re diffracting light, remember - of how much you actually know, or looked at differently, how many risk layers have been peeled off.

And because de-risking and price (understandably so) used to move together, an entire inference mechanism could rest on top of the “stage” proxy: entry prices increased predictably stage-to-stage, so stage implied check size, check size implied ownership, ownership implied reserves, reserves implied fund size (after adding the fees on top, naturally). This made “your fund size is your strategy” a logical stance, from a mere arithmetical output.

In 2026, the proxy has crashed. Ineffable raised $1.1bn merely because of David Silver’s past achievements while Arga raised $10m for an existing product with real users (early users, but still). One has already started its feedback loop with the market and is priced accordingly ; the other is still a literal puzzle but is capitalized like a late-stage industrial company. Both are labelled as “seed rounds”, which means that the same word now spans several orders of magnitude in valuation - not a reasonable 1-10x like in the 2000s, but 1-1000x. Stage and financial maturity have decorrelated.

Further evidence of that is the struggle that many established, growing companies face in raising their rounds. I won’t name names, but if you’re in business you know that many companies with real revenue, functioning teams and live operations are failing their bridges and forced to throw the towel or accept subpar deals. They are growing, yes, but apparently not enough to justify being funded again. More evidence, less capital ; less evidence, more capital: the market seems to be running upside down (5).

As a result, “we do seed” no longer means anything. It does not describe a risk, a price, a check size, an ownership target, a hold period. At best it says you’d like to be first in line. Well, as Harry Belafonte said: ‘jump in the line’, because you’re not alone there.

Size matters (6)

If we can’t derive information from the stage, then the next best candidate appears to be round size, doesn’t it? Surely the largest rounds are the most informative ones: the more money moves, the more the market must know.

In 2026, it’s actually the opposite: you can’t infer anything from a billion-dollar round! Just take for instance two of our liminal examples: Anthropic and Ineffable.

Let’s discuss Anthropic’s Series D/E (resp. $2.8bn in Nov 24 at $18.4bn post and $3.5bn at $61.5bn post in Mar 25) rather than the recent Series H (we still need to see how the IPO pans out). Back in Edition #3, we coined a term for this: Consensus Capital. Is Anthropic at this valuation a good deal? Most likely yes - and no investment committee on Earth would seriously dispute it (not even an early-stage one, as best argued by Point 9’s Christoph Janz in ‘Why We Invested in Anthropic’).
Anthropic is the poster child of Consensus Capital: it has a mind-blowing product with spectacular demand, observable performance and technological advance, an established team of world-renowned operators, an unbelievable revenue trajectory and the set of plausible winners on the LLM space narrows day by day. The initial uncertainty - does this work, does anyone want it - has been resolved. What remains is an industrial question: can enough capital secure enough compute, distribution and market position to capture a prize everyone already sees. Capital here is not fuel for discovery: it is the competitive weapon itself. You can actually understand that from who’s in the syndicate: when Fidelity, Jane Street, Cisco and hyperscalers show up alongside Lightspeed and Bessemer (the D/E rounds investors), the roster itself is announcing loud and clear that this transaction has left venture’s risk-layers peeling territory. You don’t bring a knife to a gunfight, or in this case: you don’t bring infrastructure capital to a product-market fit experiment.

Ineffable is the opposite. There is no product, no revenue, no cash flows. The deal thesis underwrites a person (David Silver) with a research paper and the possible existence of a category in the space of superlearners (one with adjacent competition at that - Jeff Dean also announced leaving Deepmind this August to work on closed-loop reinforcement learning. And he’s also raising a billion dollars). Which means the price cannot come from traditional valuation methods based on properties of the asset (7). Rather, it comes from external factors, and more specifically the scarcity of tickets available. The competitive question does not hinge on Selecting - there is nothing to judge the deal on really, besides whether David is a potential great founder - it’s all about Accessing / Winning.

Decorrelation from traditional valuation methods also brings us to the strangest property of those mega-rounds: the amount is not sized to the plan ; it is the plan. Because of the massive amount that’s raised, part of the deal’s uncertainty gets ruled out, nipped in the bud by the sheer size of the check. The company gets anointed as an “obvious winner” (an oxymoron in the investment world, if you ask me). And once declared, the alleged “obviousness” does the heavy lifting. The billion is the signal that pulls the sought-after researchers together, that flocks talent, that marshals the compute allocation, and even that commands the next round’s markup into place. The round itself manufactures a good part of the de-risking. I call it engineered success.

However the “engineered de-risking” only goes so far. Some risk layers are genuinely peeled by a large check: procurement risk (the company can buy compute), financing risk (a company with a billion of runway does not die of a missed round), talent risk (everyone wants in). Those are real risks that (used to) kill frontier projects all the time, but they are also support functions of the company (procurement, finance, HR). The core layers underneath - does the technology work, does anyone want it - are not touched by the check because no check can peel those - only building, shipping and selling does. Engineered success is the outer risk layers being peeled, while somehow the market reads it as if the core had been peeled too. The company has been made “unkillable” (or default alive by abundance of capital) but it’s not the same as being made right.

And that’s why the money, for all it engineers - beaconing talent, compute, press, etc. - cannot do the company’s actual work: building something, putting it in front of people, and finding out whether they want it. Throwing money at an entrepreneurial problem won’t solve it. That part - build something people want - remains stubbornly artisanal. And incidentally, that’s the one that will create revenue and justify the valuation (8). Capital can raise the odds of getting there but cannot substitute for arriving. And interestingly, OpenAI and Anthropic are exhibits showing that the arrival is possible, creating survivorship bias in the “Ineffables” of this world. Their success makes it look like a repeatable process, a “new normal” that lets an entire cohort of companies raise under the assumption that the actual company-building will take care of itself. It could very well be wrong, but we won’t know for a very long time, because even through crises and turnaround (such as what is going on at H Company), a company with a runway in the hundreds of millions can remain alive virtually forever.

All of that gives an Ineffable-style investment a return profile that venture has never seen before. The returns ceiling is capped arithmetically - from a $5bn entry point, doing 100x, even without factoring in dilution, would require a company in the top 30 largest ever (the half-trillion market cap threshold is that rare). Compare that to the 1,300x+ that early investors in Hugging Face just materialized. And the floor is not even the 1x liquid pref over the billion invested: no acquirer of a failed lab will pay a premium to the preference stack - if they buy, they’ll buy at pennies on the dollar. So discount on the price, plus whatever goes to the team as the usually generous packages, all of that is dollars taken away from the waterfall. So the real floor is the unspent cash which decreases by the GPU-hour. So you have the worst of both worlds: truncated upside and very palpable downside. Rational to hold only if you don’t plan to be holding it for too long (9).

Let’s compare the two: Anthropic’s investors know what the company is and what they’re paying for. The open question is industrial (whether enough capital secures the win). Ineffable’s investors don’t yet know what the company is, they are paying to be there when the team finds out, and the round itself is part of what makes the finding more likely. These are very different investments, yet both get the headline “an AI company raised a (very) very large round”, as if the size of the check told us anything about what was bought.

How dramatic is that: a billion dollars no longer tells you anything about the round and what type of risk is being financed - only that someone found the story big enough. But what’s worse is that the ambiguity does not stay at the top, it ripples down the whole scale:

• a $100m round used to mean growth: a proven company buying speed. Today it may be that, or a re-cap round for a lab that hasn’t shipped, or a team of B-lister researchers from hyperscalers that can’t be priced the same as the A-listers (the Silver/LeCun/Dean tier) above it, or a biotech buying one clinical milestone

• A $10m round used to be a Series A: Arga’s is a seed, and elsewhere the same amount is a bridge covering its name, or the ‘friends and family round’ of a repeat founder.

So size is joining stage in our graveyard of proxies.
What’s our next candidate? Well, maybe something that cannot be decorrelated from the business itself: how much capital a company actually needs to exist - how much it actually burns. Spoiler alert: that variable, it turns out, is diffracting harder than all the others.

Cost of living: everyday, we be burnin’

How much does a company need to burn to exist? This is something that’s difficult to know by essence. Indeed, it is quite different from “how much did it raise”: raising is a market outcome (how much were investors willing to pile into that business at that moment), while burning is a property of the business (how much are we spending each month in operational and capital expenditures compared to what we earn).

Because I’m not a part-time CFO at any of the companies in our list (poor me - especially not at Anthropic) I don’t have the actual financials at hand, but because I am a part-time CFO nevertheless, I think I can provide a fair assessment for each of their spending structures. And as you’ll see, the burn legend is a much better indicator in our mental map than stage or round size, because it gives us recognizable shapes.

The treadmill. Anthropic’s need for capital grows because it is winning: every new customer is a new compute bill, every top researcher it hires is another million in total comp - but it’s not just the creeping “cost of goods sold” and the lavish payroll. Critically, every model generation requires more data-center commitment - CapEx. Success is precisely what generates the next capital requirement. It’s a treadmill, a rat race, and one Anthropic willingly runs on.
Ineffable’s burn is a variant of the same shape, one step earlier, a sort of lump: the company sits on an enormous pile of cash only to find out whether there is anything there. It hires and buys compute for internal use - and it can do so for hundreds of months without worrying about the runway catching up. If the bet fails, more capital doesn’t change the answer. If there is something at the end of the tunnel, congratulations: the company graduates onto Anthropic’s treadmill and its burn becomes a function of winning (and, if we’re pragmatic, of investor appetite too: labs raise when the market is offering, not when the roadmap demands).

The staircase. Then comes the deeptechs, with a burn that’s ever-increasing. At first, the P&L is already scary for any SaaS founder - a scientific team, lab space, components, consumables, lab equipment, machinery, patents to file, insurance. Then real prototyping starts and the invoices arrive in blocks: large CapEx, assembly, iterations, more specialized engineers in the team than there are months of runway. And then the trial phase approaches, and the burn steps again: the CRO contract, manufacturing in batches, the regulatory filings, the sites (and their patients if you’re a medtech). And I’ll share with you the pereversity of the model that I know well as CFO: each success multiples the next bill. Phase I proves just enough to justify a Phase II that costs five times more, which justifies a Phase III that costs five times more again. Passing the test is what makes the check bigger. Fortunately, Venture has been financing this staircase format for decades. It is a game for specialists because it demands specialist knowledge of the ball.

The calibration: an affine function. Now we arrive at the AI SaaS (Arga), where the spending structure is finally recognizable: payroll first, always payroll - a dozen engineers at market comp - then the inference bill, then whatever is left for ads and distribution. Nothing exotic. What’s interesting here is what the size of the round implies from a burn perspective. The comp budget sets which tier of talent signs (and the round size itself does half the recruiting: almost nobody leaves a good job for a company that raised $2m). The marketing line sets how much distribution gets bought. The compute line sets how much of what they build runs on machines instead of people: in the product it’s COGS, in the team it’s opex, and the ratio between headcount and compute is the company’s operating philosophy. The round calibrates what kind of company Arga gets to be and for how long. An almost linear shape, with an extra bump (or not) when the round happens: that’s an affine function.

Zero burn. Finally we reach Medvi and it’s one where … there is no burn to model. Revenue covers everything: the ad spend that acquires the patients, the compounding pharmacies that fulfill, the payment rails, the software - and now the agents - that replace the org chart (that and an army of freelancers). The cost structure is lean and almost entirely variable: every line item scales with revenue. That’s the supreme of supreme goals in tech companies: a fully bootstrapped billion dollar business - and it’s also the reason why there is no ‘Medvi round’. A financing round exists to (a) carry spending that arrives before the income does (all our examples above) or (b) invest to supersize the business - expansion, acquisitions, push the pedal to the metal in ad spend (e.g Lovable). Medvi has arranged to have no such spending.

Which leaves a large and weird middle, where we find our e-bikes marketplace and the companies looking for bridge financing. This is where the burn structure looks more like what you’d expect to see in a regular company. The burn buys payroll and ops against a business that doesn’t yet cover its expenses with revenue: the $2m it’s chasing is eighteen months of salaries, nothing more. Money buys time - not market conquest, not proof, not a leapfrog to the next milestone, not calibration. Those are the companies that are the most ‘respectable’ in a way: they have real revenue against real opex and a gap in their pocket that needs to be filled with cash in order to keep growing. Their burn is exactly the shape the 10-year fund was built to finance. And that’s the tragedy: their need hasn’t changed one bit. What changed is everything else the same dollar can do - and the buyers of those decades-old shapes are migrating to the two ends of the spectrum.

So let’s close our mapping of the New Deals. Stage and size don’t tell you much. But the burn is imposed by what the company actually does: you can’t lie about it. And burns now run from zero to infinity, in shapes ranging from one lump sum slowly decaying (Ineffable) to a staircase (deeptech) to a treadmill (Anthropic), with the historical middle - the one honest, canonical, fundable-by-design shape - disappearing in between.

This disappearance, by the way, is no accident: it’s being propelled from both ends at once and by the same force. Back in Edition #1, we noted that building ordinary software had become trivial, pushing the residual risk into distribution. Medvi is the pinnacle of that movement: a company whose revenue scales without headcount and without capital. AI is collapsing the capital requirement of one whole class of companies - and simultaneously, at the other end, the labs and the physical-world builders (compute, energy, defense, space) need more capital than venture has ever assembled. Some companies now need almost nothing ; others need almost everything. The middle isn’t emptying because founders stopped building there - it’s emptying because AI is pulling companies out of it, toward both extremities. And that stretch is not one the same financial instrument can span.

Which leaves us with the real question of this edition: what does a financing industry do when its standard product was built for a standard deal shape that no longer exists, replaced by very different cases stretched across the spectrum? It rebuilds the product. Several products, actually - one per burn shape.

Wrappers Delight

Now let’s turn the prism toward the other end of the chain: the vehicles.

A decade ago, the landscape had essentially one vehicle in it: the 10-year €100-500m fund. Today, we have deal clubs, syndicates, deal-by-deal SPVs by dozens assembling around single transactions with no funds at all. Solo GPs running nine-figure vehicles from a laptop and a newsletter. Hyperfocused funds that only do defense, or only do bio, or only do agentic infra. Corporates and family offices going direct, skipping the GP entirely. And at the two ends of the size scale, megafunds transforming into behemoths that no longer resemble a fund while micro-funds swarm beneath them.

You probably read the reasons for where all this comes from a thousand times. Yes, it is supply-side driven:

• LP capital fragmented: pensions and endowments can’t put checks into a $10m fund, but a new breed of LPs - one that includes family offices, exited founders, operators, HNWIs aggregated through feeder platforms - can. Small vehicles that structurally couldn’t exist at scale because they had no buyers now became fundable.

• Distribution economics made a single individual fundable: X/Twitter, social media, Substack, podcasts etc. gave individual investors the same (or more) credibility towards founders and LPs than an office in Sand Hill Road with an immemorial name. The brand of the fund collapsed into the brand of the person - and so LPs started buying the person.

• The infra got cheap: all an SPV now requires is selecting among a half-dozen of providers (10))

• The 2020-2021 vintage (Tiger, again?!) made blind pools harder to raise. Anyone who was “on the road” fundraising after 2022 knows.

While I believe in that explanation, I also believe it doesn’t fully explain the change we’re witnessing. What it explains is why new vehicle shapes became cheap to try. And many were tried indeed: rolling funds, crowd-equity platforms, token wrappers, revenue-share vehicles - most already gone or marginal. But explaining variation does not explain survival. For that, let us go back to the burn shapes: the wrappers that lasted are disproportionately well fitted to them.

The lump calls for SPVs
Let’s start with the Ineffable-type deal: all the weight of the deal is on Access, nothing to select (there’s nothing to analyze), no supporting work to do after signing except wait (the founder can get their own intros easily enough). What does an LP need from an intermediary asset manager here? A way in. Not two months of deal analysis nor a decade of board work: just a ticket. The vehicle fitted to that shape is the SPV: one deal, one temporary cap-table dissolved when the position exits. Interestingly, the SPV operations come in two economic varieties. The mercenaries charge at entrance: a fee on each ticket, little to no carry - the club is a brokerage, and access is the product being sold. The missionaries charge at the exit: no fee, or next to none, all carry - the club sells itself as a conviction business and the product aligns. In both cases the intermediary is being paid for one scarce thing: the ticket. Notice what both have shed: the annual management fee on committed capital (there is no decade of work to pay for, so nobody pays for a decade), the blind pool (the deal is known before the money moves), the portfolio (the LP cherry picks their own). Every feature the classical fund built for the canonical deal shape is, on this deal type, dead weight.

The treadmill skips the GP
Anthropic-shaped deals go one step further. When the winner is visible and capital is the weapon, the intermediary’s “value-add” adds nothing that the LP doesn’t already have - which is why the LPs showed up in the round themselves. Hedge funds, sovereigns, family offices going direct: this is not a new wrapper so much as the disappearance of the wrapper, the deal itself working against intermediation. Also notice that SPVs got gradually kicked out of the party: when there’s so much demand, Access is increasingly restricted: Anthropic banned SPVs from its cap-table, other companies like Stripe and SpaceX famously have heavy transfer restrictions (11)
The only remaining “VC”-like players on this shape are the megafunds (Sequoia, a16z, Lightspeed, GC), that managed to stay there and adapt by becoming something else: permanent capital bases, registered investment advisors, multi-strategy platforms holding equity, credit and secondaries side by side. We said in Edition #3 that AUM eventually reshapes strategy, but here the strategy has reshaped the firm. Whether these are still venture firms is a fair question, one that remains priced at 2/20 though.

The affine function still needs a fund
The Arga-shaped deal is where the classical product - the fund - still makes sense. Every factor of the equation is at work: Access to get into the round, Selection because a thousand lookalikes raised the same quarter, Strategy because ownership and reserves make the fund math work or not. The $10m deal needs someone who can write $4-6m, price and orchestrate the round, take the board seat and reserve for the A. That is a $100-500m fund with a handful of partners: the most common fund type in venture.

And around the lead lives a second role: the follower. Micro-funds and solo GPs don’t price the round, they fill it based on a relationship, a specific angle, or just speed and primary access. Their wrapper is fitted to that role: a small core fund for the ticket, an SPV for the one company the GP wants more of (yours truly is involved in a very similar operation).

This shape is broad and remains the largest pool in the venture: the traditional fund isn’t extinct, but its territory has shrunk to those deals - and the deals themselves are changing.

There are two things about the 2026 shape that the old wrapper can’t absorb on its own.
First, winners now outgrow the reserves: A, B and sometimes C compress into a year for the winners, and so the pro-rata the fund would like to take is bigger and faster than the fund planned for. So it adds an extension - the opportunity vehicle or the deal-by-deal SPV - to keep buying a company it can no longer follow from its own balance sheet.
Second, a bunch of good companies need fewer rounds: some reach revenue on the seed and drift toward the Medvi end of the spectrum, which means the ownership you take at entry is most of the ownership you will ever get. That pushes the fund in the opposite direction: toward taking more ownership earlier, concentrating harder at entry, and relying less on later rounds to build its position.

The staircase calls for specialists
Deep/biotech burn in steps - each milestone multiplies the next bill. This shape has bred its own vehicles for decades:

• the venture-studio/fund: Flagship, Third Rock and others of the kind don’t invest in companies, they manufacture them - form the NewCo, own mid-double-digit percentages from day one, and fund each step of the staircase from their own reserves. Where a software fund keeps a third of its capital for follow-ons, a bio fund keeps most of it, because the staircase is inside the fund’s balance sheet. Majority ownership, inverted reserves, a portfolio the firm designed rather than found: a different product.

• the crossover funds: RA Capital, Perceptive, and the hybrids between a venture fund and a hedge fund, which exist because in bio the IPO is not an exit but another step on the staircase - a financing event that simply happens to be public. The vehicle must hold through the public line, and so it needs a structure that a closed-end fund doesn’t have. And so it grew one: evergreen sleeves with public-market capacity and a mandate that follows the company rather than the listing (12).

• The third isn’t even a venture wrapper at all, which is the point. Defense, energy, or space finance the top of their staircases through project finance, government procurement and strategic balance sheets - the other risk carriers we met in the last edition. The venture fund in these sectors is only fitted to the first steps.

Zero burn: no one yet?
Interestingly, deals that have Medvi’s shape breed no wrappers at all. They have little or no burn, therefore no structural need for a round, and therefore no obvious need for venture capital. A financing wrapper exists to carry spending that arrives before income, and this class of company has arranged to have none. If capital ever reaches these businesses it will likely not look like venture - it will look like revenue-based finance, debt, dividends-driven structures (OnlyFans’ parent company paid monthly dividends of 10+ millions to its late founder), or a holding company that buys and keeps the position for the long term, Buffett-style. Whether such a product should exist and who builds it is a question for a future edition.

The baseline gets abandoned
The remaining shape is where we find the e-bikes marketplace and other companies in need or refinancing. This was the canonical deal, what the €100-500m generalist fund was built for, and for forty years it was most of what there was in venture. That shape still exists but it no longer grabs the same share of venture attention. The generalist funds that used to finance those deals migrated - up, toward the consensus end of the spectrum where the megafund converts into a platform ; or sideways, towards the AI SaaS where the same fund now leads $10m seeds instead of $3m marketplace Series A rounds.

That shape still has financing needs, but the same fund can now chase companies with faster fundraising cadence, larger rounds, stronger LP signalling and quicker mark-ups. Its financiers found richer ground - or at least ground that currently looks richer.
And so as a result, the “standard deal” is collapsing, but not (only) because the megafund is eating the mid-sized fund. It’s because the mid-sized fund wants to imitate the megafund and abandons the burn shape it was built for, because every other shape now pays better for the same skills. It’s the frog that wanted to be big as an ox. And thus, companies are failing their bridge not because they stopped deserving capital, but because the wrapper moved on to deals that looked more promising (and that spend more).

So the two ends of our chain do line up. Held to the prism, venture capital was not really a white light, a single financial product: it was several, travelling together for as long as the deals they financed had one shape. The products are separating with them - multiplying where intermediation remains valuable, specializing where the financing need is peculiar, and disappearing where there is nothing left to intermediate. (13)

Monochromes: one colour at a time

Way before Malevitch or Yves Klein, Alphonse Allais - one of my favourite French authors, because of his cunning humour - painted monochromes. This one is called ‘Apoplectic Cardinals Harvesting Tomatoes on the Shore of the Red Sea’. The title promises a whole scene - cardinals, tomatoes, a shore - and the canvas is a single color that lets you do the work of seeing it. I’m not one that likes to explain jokes, but keep that in mind because it is roughly what has happened to the “venture capitalist” over the past few years: promises a lot, but only delivers a single colour.

What we showed in the previous section is that each wrapper we now find in VC is a form of monochrome: it sells one factor (okay, sometimes two) and sheds the rest. What we did not say yet is that the person behind the wrapper is not doing the same job either. The wrapper determintes your deals, and alongside, it defines your profession.

Let’s take the SPV and all of its cousins: deal clubs, syndicates, sidecars, collectives, networks, etc. On an Ineffable-shaped deal there is nothing to select and nothing to support, so the intermediary’s day is like a broker’s day, made of two things: getting a ticket in the round, and getting LPs to fill it. Both are sales. The first is Winning in its purest form - being wanted or being early ; the second is fundraising, deal by deal, with a story to sell every time. There’s no question of ‘investing judgement’, (hopefully) not because the intermediary lacks it, but because the deal type offers nothing for it to work on. In the mercenary variety, this even becomes explicit: fees are paid at entrance and little-to-no carry means the LP pays for the entry ticket. Some see it as a scandal, which it isn’t: it’s a brokerage. It becomes a problem only when the ticket is priced - or narrated - as investing. The missionary variety keeps skin in the game through carry, but the underlying job is the same: it is still paid for being in the round.

Making it in a round, though, is not an evenly distributed skill, and it is worth asking why one club gets a ticket in Ineffable while another only ever gets into Tier-C deals. In some cases, raising a billion upfront certainly can be akin to raising public money, and so if you’re struggling to get there, you take the closest equivalent to public money in venture: retail investors in deal clubs. In an oversubscribed round though, the founder allocates. And the founder’s criteria are ones we’ve encountered in previous editions: who the broker is (relationship, reputation, brand), what the capital looks like (already committed, able to close in days or hardly scraped together after 4 weeks of reminders), and what the ticket implies on the cap-table (are the LPs sophisticated or not, is it one line and no noise). Anthropic banning SPVs is the limit case: a company that can choose prefers no club at all. Access, for a club, is mostly built before the round.

But interestingly, a club without access does not disappear. It goes to the other end of the spectrum, to where the companies are in need of cash. There it finds the abandoned deals we mentioned: the marketplace, the bridge nobody wants to lead. And then selection reappears, but in a different form: not “is this a good company“ but “will my LPs buy this story“, “can I sell this”. Clubs with wide, non-professional LP bases sell narratives more easily - saying “we cure strokes” goes a long way for a company with trials four years away - which is why medtech and consumer narratives do well there and open source infrastructure doesn’t. Some clubs are at least honest about it: they set a low threshold and see themselves as the company’s fundraiser. The fee schedule also reflects it: at the top of the scale the club is paid by LPs, for a ticket. At the bottom it is paid by the company as well - a success fee on the raise - and by the retail investors on top of it. Fees are fees, on both ends (14).

So the deal club is one wrapper with two mirror-image economies: a broker of access when capital is abundant and the founders picks, a broker of capital when access is abundant and the company needs. It is the same sales job in both cases, what changes is which side pays you to find the other. Notice too what this does to the abandoned deals: it does not totally stop being financed, but it gets re-intermediated at a worse price - a placement fee, carry and a cap-table with a three-hundred-heads SPV that the next lead will not want. That is the second blow of the mid-sized fund’s migration on the e-bikes marketplace.

There is also an LP’s side to the deal club story, and to me it’s an indicator of how sophisticated LPs are getting when it comes to venture. Seen from the LP’s chair, a deal-by-deal ticket at Series C or D in a household name - Anthropic, Databricks, Cursor, Mercor, your pick - should not entirely be seen as a venture bet. It is a “safe and predictable” 2-3x, liquid at the E or F round some eighteen months later, when the hedge funds, sovereigns and family offices we mentioned earlier arrive direct and the secondaries that the round begets take the position off their hands (15). Yes, the upside is capped, as we said of Ineffable, but the LP does not need the upside: it needs the next markup to happen. Compare that with the fund product. Below the 95th percentile, the benchmarks of Edition #2 return less than that multiple, take a decade to do it, and lock the money for the whole run. Do the IRR math: 2.5x in 18 months is ~84% while 2x in 10 years is 7% (if your fund manages to do it). So the ticket is not selling access alone, it is selling short duration - a (good) venture multiple on a holding period that seems like a bond. The catch is where the return comes from. It does not come from the company, rather from the sequence of buyers: the SPV holder at Series B is paid out by the direct LP at Series D. As long as the queue moves it is the best product venture has ever offered an LP. When it stops - 2022 will do as a reminder - the ticket becomes an illiquid line at a zesty price (9 - again).

I went deeper on the SPVs because that’s where the illustration of how the job changes is most vivid (and also so I could place the pun ‘fifty shades of prey’, which I couldn’t so now here it is), but it’s true for other wrappers, too:

• The studio is another monochrome, indexed on Supporting. It doesn’t pick among companies: it builds and funds them. The factor they sell to their LPs is the build: Supporting pushed to the point where the firm is the operator. The people inside are scientists and company-builders, sometimes the founding CEO.

• The platform - the megafund that became a registered investment advisor with public equities, credit and secondaries sleeves - sells something else: an unlimited balance sheet. The job is asset management: raising, allocating across strategies, keeping AUM in motion. That it remains priced at 2/20 is remarkable.

• This leaves the traditional fund as the only wrapper still selling strategy, access and duration together - all pillars engaged on the few deal shapes that still require them. In the old sense of the word, it is the only one whose manager is still a venture capitalist. Around it, the follower - micro-funds, solo GPs, yours truly - sells a relationship, an angle or speed: access again, with much smaller footprint.

We had a section in an earlier essay called “what you do is who you are”. This is also true in the reverse here: “venture capitalist” now designates a broker, a builder, an asset manager or an investor. Wrappers should be monochromatic: they can only contain one flavour because it implies a different job. And there is nothing wrong with a monochrome, but still collectors are tempted to juxtapose colours to create beauty. Which raises a question that a firm, as opposed to a wrapper, has to answer for itself: how many colours can one firm hang on the same wall?

Qui trop embrasse mal étreint (16)

Should a regular fund do Consensus deals - or Access deals (Ineffable), or the odd deeptech ? This is a question with two answers, at different levels (wrapper and firm).

At the level of the wrapper, the answer depends on the shape - which is the point of this whole edition. On a lump, it is no: an Ineffable ticket in a $200m fund charges its LPs 2/20 and a decade for Selection and Strategy the deal does not use, and the same LP can buy the ticket next door, from a club, without being locked for a decade. On a treadmill it is less simple, because the club is barred (Anthropic bans SPVs) and the LP is either large enough to go direct or has no way in. Which brings us back to the Point 9 parenthesis of earlier, which really isn’t a parenthesis any more. Summarized: the German early-stage fund took a position in Anthropic’s Series G at $380bn post-money. It is a good deal: first, the ticket was not chased but inherited from a portfolio company (Vercept) that was acquired by Anthropic. The position rolled over with a right to double down. Second, the rationale was pure consensus: run-rate from $1bn to $9bn in a year, exceptional product, etc. Third, it was declared as an honest exception: not shifting the firm to a late-stage fund and not a new strategy. The math stays what it was - from $380bn no ticket returns a seed fund. The line is there for a near-certain 2-3x and, come IPO, for DPI years before the rest of the portfolio can produce any, which we said earlier is the one number LPs actually track. That is a legitimate thing to hand your LPs, as long as you say that is what it is. A wrapper does one shape. When it does another, it should be able to say why.

And by the way, the treadmill SPV stint also works in the old wrapper: Factorial Funds started as another deal-by-deal competitive rounds SPV operation until the manager seems to have decided that the brokerage was too exhausting and raised a $135m fund in 2022. As of 2026, the fund sits at 7.1x DPI (!!!), while the median for the vintage is 0.0x. The portfolio is made of Consensus deals only: Anthropic, Anduril, xAI, Perplexity, Mistral, Eleven Labs, SpaceX, Thinking Machines, Physical Intelligence, etc. In this case, you only sell Access, and the wrapper still does one shape only.

At the level of the firm, the answer is different, because a firm is not a wrapper: a firm can have several wrappers. This is the natural evolution of most funds: the core fund is surrounded with opportunity vehicles and SPV programmes beside it, RIAs have evergreen, credit and public equity sleeves. I co-head a small vehicle that happens to do deal-by-deal too. So the real question is not “should the fund do it” but “what does it cost a firm to hold two colours”.

• Cost structure. Each wrapper comes with its own cost base. The fund’s fee pays for people who select and sit on boards for years ; the SPV pays per ticket and needs a sales machine (memo, data room, KYC, investor CRMs, and lots of calls). A firm holding both carries both, and so the fee of one ends up paying for the work of the other: the fund’s management fee funds the partner-hours spent placing SPVs, or the SPV fees keep a small fund’s team alive. Either way one set of LPs subsidises another.

• Fee schedule. Two products need two prices. You have a kosher version where fund LPs get the SPVs fee-free or at reduced carry, as a co-investment right and outsiders pay full price. And you have a messier version that double-dips: the fund keeps charging on reserves it does not deploy because the follow-on went into an SPV … that charges again. Fair if you have no reserves, more difficult to explain if you do. Sure, it’s a win-win, partners get carried faster, LPs get IRR. The fiduciary problem occurs for the LPs that did not go direct and subsidise the others: that follow-on could have been in the fund, and now it’s not.

• AUM. The two wrappers pull it in opposite directions. The fund’s size is set by its strategy - ownership, reserves, burn shape - and takes years to move. The SPV programme’s size is set by what LPs will buy this quarter, it can easily double in a year. We said in Edition #3 that AUM reshapes strategy. Well, inside a multi-wrapper firm this is the channel through which it happens. The deals the SPV operation can place - legible, consensus, with a mark-up in sight - start to look like the deals the fund should lead, when really it shouldn’t. Strategy gets blurred, AUM pulling it towards different deals.

• Allocation. When a scarce ticket comes in, which wrapper gets it? The fund’s LPs paid for access and judgment together ; the SPV’s LPs paid for access alone. Pro-rata to the fund, surplus to the SPV, is the defensible rule. The other option is to give best access to the SPVs because they pay per ticket, but it turns the fund’s LPs into the subsidisers of a brokerage operation. The allocation policy is consequential in a multi-wrapper firm’s documents.

And then there’s another input that no wrapper scales: partner attention. We just showed that each wrapper is a different job. A firm with three colours asks from the same three people to be tour-à-tour broker, investor and builder in the same week, and to switch between meetings. The power law of people from Edition #3 does not work like this ; the craft of Edition #5 does not work like this either: the firm thus gets good at the colour that pays fastest and mediocre at the others. Qui trop embrasse mal étreint is a reminder that your attention span is a limited budget as a VC.

So a firm can hang several colours of the monochrome, and in 2026 it probably has to if it wants to survive. What it cannot do is mix them on one canvas. Each wrapper stays one colour, priced on its own, with a written rule for who gets the ticket, and a fund whose strategy does not depend on what the SPV operation can do or what its studio is building.

The frame hasn’t moved

Let’s finish by highlighting something we said at the beginning of this article: every deal in our opening list shares one contract type (preferred stock or a SAFE). As I noted in footnote 2, I borrowed the word “diffraction” from my former colleague Nicolas, who uses it not for the deals or for the vehicles, but for the financial products themselves. Deals diffracted, wrappers too, but the contract has stayed the same. Every monochrome in this section still hangs in the same frame: a decade-old template from Y Combinator (or a BSA AIR if you’re in France) negotiated by the same lawyers with more or less the same term sheet.

But let’s have a proper look at what the same financial instrument really means in our list:

• On Anthropic, preferred equity is one step away from public equity: a Series H with a listing pending means that the preferred equity is juste a temporary formality - it will convert to common upon IPO. And you might as well call it liquid.

• On Ineffable, the same preferred equity resembles a call option on a research outcome - the 1x preference protects nothing (footnote 9) and the company is worth zero by the book (footnote 8), which means that what the investor bought is a claim: the right - not the obligation - to buy more if there is something at the end of the tunnel. This is the textbook definition of a call option (and I have my Vernimmen open)

• On the deeptech staircase, the equity is a tranche. The next is released when a milestone is met. We already argued that the top of the staircase was replaced by project finance and government procurement, well it’s probably because that is what it is in the first place.

• Our AI SaaS is the only “classical thing”. Preferred equity meant for financing venture-type risk.

• On Medvi and other AI-bootstrapped companies, there is no equity at all. There might be some models we already mentioned: revenue-based finance, dividends-first, holding companies (à la Constellation)

The equity instrument has not diffracted in form but it has diffracted in meaning. The same contract describes an option, a project-finance tranche, quasi-public stock and a regular seed round.

Deals move when founders and markets decide ; wrappers move when a GP and a few LPs decide ; contracts move only when everyone decides at once - founders, funds, lawyers, regulators. That’s why the contract is sticky. But the gap between form and meaning does not hold forever: just like the wrappers decluttered and removed the parts that they weren’t using ; someone somewhere will iterate with new instruments, fit to the the new deals. Because it’s all survival of the fittest, there might very well be one that sticks. And that’s why the largest firms have already started to transform into RIAs: so they can do it all, by design and by law.

Thought #7: pick your poison

My working thought all summer, scribbled before any line of this article was written, was “pick your poison: what deals, thus what wrappers - or the opposite“. It was a surface-level assumption at the time: you can only be good at the thing if you have the right vehicle for the thing. Having now held the light through the prism, the thought holds - it just needed the essay to understand what the poisons are.

Because each of these products has an underlying poison: none of the colours comes free.
The SPV gives up judgment and capital availability for flexibility. The fund gives up liquidity for a decade and diversification of deal types (it has to pick one strategy). The studio gives up doing external deals. The megafund gives up being a venture firm to turn into an asset manager.

There is no perfect option on the shelf, so a poison will be drunk either way. The only question that matters is whether the LP knew which one was in the glass.Said differently: the crime, in everything we’ve covered, is never the product. It is the divergence between what a manager sells and what a manager does.

Every case we mentioned in this edition can be sorted by this single test.
Factorial sells access and delivers access: a blind pool of tickets into names its LPs could not reach otherwise, upside capped, exits whenever the multiple makes sense - and the fund wrapper is not dead weight there, it is the access itself, because committed capital is what closes in days and the hottest cap tables ban SPVs, so the “ticket next door” does not exist for an Anthropic or a SpaceX. What’s sold and what’s done coincide; the 2/20 pays for a capability the deals fully consume. Point Nine sells a decade of early-stage judgment and delivered exactly that, with one Anthropic line on top, labelled as the opportunistic exception it was. Coincidence.

Now the divergent cases. A fund that sold its LPs judgment and that keeps deploying into deals where no judgment went: they pay for machinery that never turns. A firm whose SPV operation ends up pushing what the fund invests in: the LPs bought one strategy and are served another. The mid-sized fund that leaves the deals it was built for to chase the megafund’s: the same divergence, at the scale of the whole market. None of this is fraud, and most of it is not even bad faith - it is what happens when the same term, the same financial instrument and the same fee schedule get stretched over products that stopped being one thing. But the gap is always paid by someone, and in this edition we realized it was mostly paid by the companies of the middle, the abandoned deals whose shape didn’t change while their financiers’ incentives did.

So the thought stands as written. Pick your poison - deals first or wrapper first, whichever way you come at it. Just pick one, name it to the people drinking it, and charge for that one only.

Subscribe now

Footnotes

(0) Some of you would likely categorize Anthropic as a Growth deal, or a Late Stage deal, possibly an Infrastructure deal. I've always seen VC as a an entire category spanning those three, from early stage to late stage ; but also, that's the whole point of this edition: the lines of categories blur.

(1) I do not know the company and am unaffiliated. It was just the first to pop when I opened LinkedIn, so I let serendipity do its thing.

(2) The term is borrowed from my former colleague Nicolas Colin, who uses it to point at different financial instruments used in VC, which is yet another layer. We’ll leave it untouched for now but might encounter it later on as a conclusion drawn from what happens at deals and wrappers layer.

(3) That shape was never unique - biotech and growth equity have been packaged in the same type of vehicle for decades. But each with their own predictable shape. What is new is not that several shapes exist, but that they now sit orders of magnitude apart and keep moving away from each other while bearing the same name.

(4) I recently went down a rabbit hole about how maps were made before cartographers had reliable instruments to measure distances or height. They were a bit like police caricaturists: brilliant at reconstructing a whole from fragments - inferring for instance which way a river must flow from the stories explorers brought back - but also prone to going very wrong, very fast.

(5) ‘In God we trust, all others bring data’ says a well-known quote that’s often found in cheap T-shirts that seem like they’re just out of a Big Bang Theory episode. A formidable thing about VC is that at the earliest stage we have no data to use (as pointed out in previous editions), so we must choose to extend trust to the founders. And the reasons to do so can be who they are (their expertise/experience). This is not something new. What is new is the length at which (blind) trust extends relative to the difficulty of having satisfactory data.

(6) Not the verb, you cheeky reader, the noun!

(7) Bidding wars are nothing new, as we’ve seen in previous editions (esp. #5). But the price a bidder was willing to pay was usually dictated not by his sole intuition and power/ego plays but also on case sensitivities. The spreadsheet gave a first indicative pricing idea to follow or dismiss. Here, they don’t even exist, you’re pulling a price out of … thin air - not to be impolite.

(8) My first ever real job interview was at Lazard Frères on Boulevard Haussmann for an internship in the TMT M&A team. And the very first question I was asked (besides whether I was ok with sleeping strictly less than five hours per day for six months straight) was: you open a company and deposit a million in initial capital - we were lightweights back then - what is the company worth? The answer, the correct financial accounting one at least, is zero (shareholder’s equity is 1m, net debt is -1m, so total balance sheet is zero). A company with cash in the bank and no other assets or activities is worth, in theory, zero. That is, if your name is not David Silver, Jeff Dean or Yann LeCun. I got the answer but didn’t get the job anyway.

(9) I wouldn’t dare say Ponzi, but I’ll say this is a wrapper problem (look further in this edition) and an exit/liquidity one (look in the next edition).

(10) My first firm was the one that imported SPVs in France. I used to have colleagues that had to manually incorporate, register, open up bank accounts for and administer each one of them for ~€100k-ish deals, leaving room for errors and other kinds of misbehaviors I won’t go into ; entities that I later had to oversee, taking over not only the deal itself, but also the admin/accounting/tax/IR part. I am first-hand grateful that the barriers to an SPV have dropped. I’m only mildly worried about the consequences on the asset class.

(11) And that’s why some managers of SPV operations venture in the questionable business of two-layered, three-layered, sometimes four-layered SPVs. I wouldn’t touch those with a 10-feet pole, but to each their own …

(12) Second time we encounter this crossover private-public format in this edition (RIAs were mentioned above). Keep that in mind for next edition :)

(13) One feature, though, has proven strangely hard to shed across every one of them: the clock. The SPV still needs its position to resolve ; the calibration fund still promises a horizon ; the crossover holds through the IPO but still owes a distribution ; even the permanent-capital platforms answer, eventually, to someone who wants their money back. The market has redesigned everything about the wrapper except the part where the money comes out. That part has a story of its own - and it’s the next edition.

(14) I suspect the two economies are often the same business seen at two moments: one hot allocation is the marketing that builds the LP base, and the LP base is then monetised on a steady flow of deals that are easier to get into. I have no data on this, only the fee schedules which seems to indicate that some of those clubs are having … ahem, a blast, with how far you can take the model.

(15) Probably picked up by another deal club, too.

(16) French saying, literally “who embraces too much doesn’t hold much” and close to “grasp all, lose all”. Possibly I should have considered it before embarking in this writing journey.

If you missed it, you can read:

• #1 - Financing Risk : a brief history of Venture Capital

• #2 - Chance, the Wrapper: VC as a Financial Product & Asset Class

• #3 - Sum of the Part(ner)s: Why VC Partnerships Fail

• #4 - Backing Great Founders : The Biggest Cliché in VC

• #5 - The Craft: How VC Firms build edge

2027 will be the year we have a new president in France, after a decade of Macronism.

To me, who usually takes a pragmatic stance regarding the State’s role in Venture (‘we’re microeconomic agents: the entrepreneurs we pick work around the constraints that States create’), I figured that this was still a naive approach since I had lived almost all of my professional career under a single regime and that it would be a good to use this series as an excuse to interrogate a regime’s change on my industry. Put differently: what the impact of politics was on VC and its ecosystem.

And incidentally, the firm I started my venture career in had a rather radical stance on the government’s role in the founder’s business, that one could summarize as: “State money bad for founders. State doesn’t understand startups. State money create zombie startups that should die faster. Don’t take State money, make revenue”. So my first opinion on the mingling of the State in entrepreneurial affairs was: it shouldn’t.

Of course, life is not that black or white, as I learned over the years.

First, I met countless founders who managed to not live off subsidies but took them for what they were (to them): a mechanism that increased their runway. At the individual scale, it always makes sense to chase the subsidy if it doesn’t suck up too much of the founder’s time - and I learned to see the large fishpond of “non-dilutive fundraising advisors” not only as a parasitic class living off poorly allocated government capital, but also as timekeepers for pragmatic founders.

Second, I had the chance to work alongside the now French Minister of Foreign Affairs (Jean-Noël Barrot), who was then an MP and a teacher of Entrepreneurship at HEC Paris and MIT. He showed me that individual politicians could have a thorough understanding of what entrepreneurship was like, and truly aim at changing regulation their way: Alice (the CEO of my previous firm whom I worked with on this matter) and him were tasked with preparing a report on how to finance companies better for the Minister of Economy (I wrote about it here). This also taught me that the boldest ideas often end up watered down in the political cycle. Policymaking is about consensus, and consensual and bold rarely pair well.

Third, and perhaps more importantly, I came to realize that the Venture Capital industry, in any country, wouldn’t have been what it is without the State’s presence. We already talked in Edition #1 about Frederick Terman and how the government was key to subsidizing Silicon Valley through demand. Yes, it’s a reverse read of what really happened (build startups corresponding to government demand, so they’ll buy the tech), but the flow of dollars remains the same: from government to the startup world. And in Europe of course, between one-in-three and every other euro in VC comes from a government-sponsored entity.

All of this made me more nuanced regarding the particular relationship of the State to VC. But nuanced about what, exactly? Rereading those three realizations, it became clear to me they aren’t three anecdotes about the same thing. They are three different things the State does. In the first, the State plays the role of capital: it extends a founder’s runway. In the second, it plays a regulatory role: it writes the rules of the game. And in the third, it plays customer: Terman’s students didn’t need venture capital because the Army was buying ; and the EIF is the main LP of European funds (their customer of sorts, if you remember Edition #2).

Why would any of this matter to a fund, though? Let’s go back to Edition #1 where we said Venture Capital is a machine that finances ventures with low odds of success, on the condition that the investment hinges on one dominant uncertainty. One: technology or market. Everything else - the founder finding customers who can grow into large contracts, the engineers being trained somewhere, the company being sellable one day, the law not changing under its feet - has to be carried by somebody, and that somebody is not the VC. We can take the risk knowingly, but we don’t carry it.

This is, I think, the definition of an ecosystem I can offer as a personal take: a VC ecosystem is the set of actors who absorb the risks that Venture Capital cannot or will not. Not a checklist of who or what is in the ecosystem (e.g journalists, universities, incubators. Sometimes you’ll find capital, talent, culture …), but a ledger of who carries what risk. And since the State is by far the largest - and the most interesting - line in that ledger, this is where we’ll spend most of our time before turning to the others.

Government-backed Capitalism

It’s always been funny to me how this industry - esp. In Europe where I’m more familiar, and myself included naturally - talks about “the State” with lots of imprecision.

The word covers a budget line at Bpifrance, a tax article buried in the Code Général des Impôts, or a procurement officer at the DGA/Pentagon - three things that have nothing in common except a passion for administrative duties and documents too long to read in one sitting. So before being adamant whether State money is good or bad for founders, it’s worth being precise about what the State actually does when it shows up in a startup’s life. There are three ways the State ‘intervenes’, three “instruments” of sorts - and I’ll argue they map onto different risks.

The State can buy. It can be the customer - the first one, the largest one, occasionally the only one. When it does, it absorbs market risk: the question of whether anyone will pay for this thing is answered before it’s asked.

The State can fund. It can be capital - through grants, subsidies, tax credits ; sometimes as a direct equity investor in ‘national interest startups’ or more often one level up, as the LP that anchors the fund that writes the cheque. When it does, it absorbs financing risk: the company gets more time (following Ben Franklin: time is money).

The State can rule. It can write the framework within which the other two happen - who may own a patent, how a stock option is taxed, whether a pension fund is allowed to hold venture at all, whether a ride-hailing app is legal. When it does, it “absorbs” regulatory risk, or manufactures it, depending on which side of the pen you’re on.

The three are usually lumped together as “public intervention”. But they behave nothing alike once you look at what each one does to the investment case, and the difference between them will end up explaining most of what I believe separates the Valley from Europe.

Demand: to the moooon!

Let’s start with the instrument that never looked like State money but very much is so: the State as customer.

We’ve already met Frederick Terman and the Army in Edition #1, so I won’t rerun the whole tape. But there’s a number I left out of the first essay (because I hadn’t read it yet - this is very much a ‘write-as-you-go’ process). In the early 1960s, the federal government bought almost the entirety of the integrated circuits produced in the United States - the Minuteman missile program and Apollo between them consumed the bulk of the output of Fairchild, Texas Instruments and their peers for several years (the sole Apollo program consuming almost 60% of the US production). Silicon Valley’s first product had a single client, and that client had an unlimited budget and a deadline set by Kennedy (1).

There was no market risk to underwrite: the market was the procurement office of NASA and the Pentagon. What remained was the question of whether the chips could be made reliable enough to fly to the Moon - a technological uncertainty, one, exactly the kind of bet a venture investor can take. Which is why, as we noted in the footnotes of Edition #1, the first startups of the Valley did not need venture capital at all: they were financed by corporate money (Arthur Rock pushed the Shockley traitorous eight to Sherman Fairchild, an industrial businessman) and when the State is the buyer, the deal has already collapsed to one risk before any investor shows up.

This wasn’t a one-off of the Cold War. The pattern is so recurring that it has its own literature - Mariana Mazzucato’s The Entrepreneurial State being the reference. The internet is a DARPA project (known as Arpanet). GPS found its way (hehe) through the Department of Defense. Siri was spun out of SRI’s CALO program, funded by DARPA. And since 1982 the SBIR program (funny name, if you’re French) has handed out a few billion dollars a year of what look like research grants but functions like first purchase orders - Qualcomm is an alumnus. The CIA even has its own venture firm, In-Q-Tel, which invested in Palantir in its early years and in Keyhole, the company that became Google Earth: the State as VC and customer at the same time, a full-stack risk absorber.

An obvious and more modern example, though, is SpaceX. In December 2008, the company had just flown Falcon 1 successfully on its fourth attempt, after three failures that had burnt through most of Musk’s PayPal money (by his own account it was weeks away from running out of cash, and Musk was not the richest man in the world at the time). Then NASA awarded it a $1.6 billion Commercial Resupply Services contract, which turned a rocket company with no customers into a rocket company with the largest customer in the world booked for a twelve flights backlog. Nothing about the technology risk had changed overnight. What had changed is that the market risk had been removed, and once it was, private capital did what private capital does when only one uncertainty is left: it poured in.

Anduril is the same story on a shorter timeline - founded in 2017, embedded in Pentagon programs and border surveillance contracts before anything an investor would call product-market fit. Sinéad O’Sullivan, whom we’ll meet properly in the next section, calls this “selling shovels to the miners”. In fact it’s even more specific than that: it’s selling shovels to a miner with unlimited pockets.

However, here is the important highlight in my view: a subsidy and a procurement contract both move dollars in the same direction: from the State to the startup. But they are not the same thing at all. The subsidy arrives as runway and leaves the core founding interrogation intact: will anyone pay for this? The contract arrives as revenue and answers it. One extends the time you have to find out ; the other tells you. In the ledger we’re building, State demand doesn’t just help the company along - it removes an entire category of risk from the investment case, which is precisely what turns an unfinanceable bet into a VC-bankable deal.

At least that’s the pitch, but I still need to raise three caveats:

First, State demand is not a market. It is scattered, political and comes with its own calendar: budget cycles, elections, the infamous “valley of death” between a prototype contract and an actual scaled program. You have swapped market risk for political risk - and political risk, unlike market risk, can be lobbied. Hold that thought for after.

Second, and this is where VC re-enters the picture: government demand builds the first product, but it rarely builds the company. Fairchild needed no venture capital to sell chips to NASA ; it needed it - or rather its offspring (Intel) did - when the chips had to find civilian markets. Equity risk capital only became necessary when startups had to find customers instead of receiving them. Which means the State-as-customer doesn’t replace the VC ; rather it hands the VC a more financially stable asset and a later problem: the dual-use pivot, from a buyer that keeps on buying to buyers who don’t. Interestingly - and unsurprisingly to my readers I suspect - a lot of the startups in Deep Tech I talk to these days solve the first caveat (time) by embracing the second (dual-use). They seek customers first while they manage to get government contracts.

Third, and in connection to the above: the State only buys what the State needs. There is no procurement office for a consumer app or a B2B SaaS - nobody at the DGA is waiting for a better CRM - so the instrument isn’t available to the industry, it’s available to a category: sovereign missions. Defense, space, cyber, energy, health. NASA removed SpaceX’s market risk because NASA needed rockets ; it would have done nothing for Airbnb. Which means that a fund building a thesis around State demand isn’t only choosing a risk-absorber, it’s also choosing sectors. Everyone else still has to find a customer the old way.

The next instrument is capital.

Capital: welcome to Zombieland

In fairness, what my old firm objected to was a caricature that very much exists.

In France: a founder chasing Bpifrance’s aide à l’innovation, the CIR, a JEI status, a regional grant and a Horizon Europe consortium, spending a third of their calendar filling forms designed by people who have never shipped anything, going from pitch contest to pitch contest, and calling the result “non-dilutive fundraising”. I’ve seen it. But I’ve also seen it work, which is what made me revise the doctrine rather than discard it.

The instrument is simple to describe: the State plays the role of capital. Sometimes directly - grants, subsidies, tax credits, subsidized or deferred loans, and direct equity stakes from time to time - and sometimes one level up, as the LP that anchors the fund that writes the check. In Europe this second form is the elephant in every data room: the EIF, Bpifrance, the British Business Bank, KfW and their regional cousins are the reason why 36% of European VC capital comes from government-backed entities vs 4% in the US.

The Americans tried it too, however, and earlier than you’d think: the Small Business Investment Act (SBIA) of 1958 turned the federal government into a lender to licensed venture firms, and for a decade SBICs (SBI Companies) were the venture industry - Draper & Johnson started as one, and Apple, Intel and FedEx are commonly listed among SBIC-backed companies. The program faded in the 1970s under the weight of leverage, fraud and mediocre returns, at exactly the moment the limited partnership and ERISA took over (head over to Edition #1 for more details). Which is a first hint: the US State tried the capital instrument before the rules instrument, and it’s the rules that built the asset class.

In the ledger, capital absorbs financing risk. Put more plainly: it buys time. A procurement contract answers the founding question - will anyone pay for this? while a subsidy postpones it. It arrives as runway, not revenue, and it leaves the company exactly as uncertain as it was, only better funded. The pragmatic reading I came to hold is correct at the level of a single founder: more time to find out or more money to spend is strictly better than less. At the level of the population, though, it produces zombies: companies that should have died and instead survived on public runway, with a team, a product and no market, then died anyway, later and more expensively.

There’s a second-order reason this happens. Subsidies are allocated on criteria, and the criteria cannot be “are there enough paying customers?” - tautologically, because if there were, there would be other ways to finance the business. So they are proxies: R&D headcount, patents filed, TRL (Technology Readiness Levels), jobs created in the right département. The instrument selects for legibility to the State, not for demand - and any selection criterion that isn’t demand eventually pushes the industry to optimize for it. That’s what the “non-dilutive advisors” are: a conversion layer between the State’s proxies and cash. They are parasitic when the founder lets them run the calendar and useful when the founder treats them as timekeepers. Either way, they are the visible symptom of a risk being deferred rather than removed.

Now let us go one level up, because the same mechanism exists at the GP layer. When the EIF anchors a first-time fund, it absorbs the GP’s financing risk - the equivalent of the founder’s runway is the fund’s close. Edition #2 said LPs are the VC’s customers ; the State as LP is therefore the State as customer of the fund, and it removes the fund’s market risk (can I raise?) the way NASA removed SpaceX’s. Except that the EIF, unlike NASA, is not buying a product it needs. It is buying proxies tied to political agendas: geographic mandates, ticket sizes, ESG frameworks, team diversity criteria. So Europe partially has a population of funds that exist because a public LP signed, not because a private one believed - the exact “partnership by default” we diagnosed in Edition #3, with a public anchor as the default. Zombie funds, deploying for ten years on the management fees paid by European taxpayers.

Is the instrument ever right, then? Sure it is. When the dominant uncertainty is technological and slow - a molecule, a reactor, a new fab process - time is the risk. There is no market question to defer, because the market is legible ; there is only a clock the company runs against, and a subsidy that buys X years of it is absorbing exactly the risk that needs absorbing. This is why the CIR, for all its abuse in software, is a good propeller for biotech. The rule follows: capital is the right absorber when the bottleneck is time on a technology bet, and the wrong one when the bottleneck is demand. Deferral is fine when deferral is what you need.

Which retroactively explains the doctrine I started with. “State money bad, make revenue” was a rule for software startups in the 2010s, companies whose only real uncertainty was market. For them, a subsidy defers the one question that matters and only a customer answers it - so “make revenue” was right, not as a moral stance but as a risk-allocation one. Extend the same doctrine to a founder trying to make fusion work and it becomes nonsense.

The same question also deserves an answer one level up, because the State-as-LP has a legitimate case too, structurally speaking. The American LP base exists because Americans fund their retirement by capitalization: there are pools of pension money to tap in, which is what ERISA unlocked (as a reminder of Edition #1: the ‘prudent man’ rule explicitly allowed pension funds to invest in PE from 1979 onwards). Most of continental Europe runs on pay-as-you-go (retraite par répartition). There is no CalPERS to persuade to invest in your fund, because there is no CalPERS - the money that would have been a pension fund is largely transferred from current workers to current retirees. Add Solvency II, which historically made unlisted equity extremely capital-intensive for insurers (a 49%+ symmetric-adjustment capital charge, versus 22% for the difficult-to-access LTEI Long-Term Equity treatment. In layman’s terms: for every €100 an insurer invests in unlisted equity, the regulator could require it to hold €49 of its own capital against the risk), and the private LP base that the Valley took for granted simply doesn’t exist here at similar scale. So the European State as anchor LP isn’t (only) absorbing a risk the market has declined to take, it is also standing in for a market that can’t form. That’s a form of bootstrapping. The thing with bootstrapping is it needs to stop doing the leg up at some point.

The good version of bootstrapping is Israel’s Yozma in 1993: a $100 million public fund-of-funds that would only invest alongside a foreign private VC, and that gave the private partners the option to buy out the government’s stake at a fixed price within five years. Every one of the ten funds it seeded was bought out. The State co-signed a private investor instead of replacing it and had written its own exit into the term sheet. Within a decade Israel had a private LP base and the program was gone.

Compare with the EIF, which has anchored European venture since the 1990s and shows no sign of leaving - not entirely by choice. Israel had a deep foreign LP base to hand over to ; Europe has twenty-seven home-biased markets in which, by the ECB’s own count, cross-border LP commitments barely exist outside the EIF’s own network. The anchor became the plumber working the pipes. And the private LPs who should have replaced it were never allowed to: no ERISA (quite the opposite: the European equivalent, the IORP directive caps pensions exposure to PE) and a Solvency II framework that has priced insurers out (until 2027). So the fund’s market risk - will anyone private buy this fund? - has been deferred for thirty years, not because the State wanted to stay, but because the rules never let it go. Again, deferral is fine when deferral is what you need, but it’s fatal when it’s the only thing you’re allowed to get.

So the rule at the LP level is the same as at the founder level, with one extra clause: the State-as-LP is the right absorber when the private LP base is structurally missing, provided it co-signs rather than substitutes, and leaves. Which, incidentally, is what France’s Tibi program tried to do from the other end - not by writing cheques but by nudging insurers into the asset class. That’s not the capital instrument, that’s the third one.

Policy: the (in)visible hand

As part-time CFO (an occupation that allows me to live off an €8m fund alongside 2 senior co-founders), a good chunk of my week goes into BSPCE plans, insurance policies, salary treatments - none of which exist because the founders or I decided it. They exist because at some point, somebody in a ministry wrote a paragraph into their bill proposition. The State can buy, the State can fund, and the State can also just … write the rules.

We’ve actually already met this instrument twice in Edition #1: the 1978 capital gains cut and the 1979 clarification of ERISA’s prudent man rule is what turned US VC from a cottage industry of a few hundred million a year into a few billions under five years. Let’s look at it again through the lens of this essay: it let the American State stop being the LP. The SBIC program from the 1960s got retired because a rule had made private LPs possible. The State exited, but kept supporting VC with the policymaking instrument: Bayh-Dole in 1980 let universities keep the patents on federally-funded research (enabling commercialization by private US companies). NSMIA in 1996 loosened the rules on private fundraising, QSBS exempted small exits from tax gains, carried interest is taxed as a regular capital gain rather than a bonus. The list of rules that apply to VC goes on and on.

What risk does this instrument absorb? Regulatory risk, obviously - but I think what’s really interesting is how differently it behaves from the other two. Capital is a budget: it has to be voted every year. It competes with hospitals, education, law enforcement. Demand is a contract: it lives through its procurement cycle. A rule, once it’s passed, keeps working for decades. It does have a cost in some cases though: a tax break means lower tax income which means lower budget next year. In France, public support to business innovation went from about €3bn a year in 2010 to close to €10bn today, and two thirds of it is the CIR alone - roughly €7bn a year of what is technically a tax rule but behaves exactly like a subsidy (2). Still it remains the cheapest tool the State has, but also the slowest one to obtain, and by far the one with the longest half-life. Which is precisely what I learned working with Jean-Noël Barrot and Alice Zagury on that report: the ideas were bold but the output was consensual, and it took a long time to accept that nuance is the price of longevity. What is hard to pass is also hard to undo.

France, of course, has its own history with rules. The BSPCE was created in 1997 and let’s say that … two reforms in and even though it’s gotten much better, I’m still having the below conversation with founders on a regular basis (3).

But beyond employee equity, French laws impacting VC are numerous - JEI, PEA-PME, minibons, IR-PME, Monetary Code, Pacte Law, the régime Arthuis: rules, all of them, all late to the party, all diluted somewhere between Bercy and the Assemblée Nationale (4). The thing to note here is that regulatory risk is one that the people who bear it can organize to remove. You can’t lobby a customer into existence, but you can absolutely lobby a tax article out of it.

And at this point, we should note how integrated this is: the three instruments do not exist independently, rather they form a coexisting system. Rules decide whether the two others are needed at all. The American State used a rule to make itself unnecessary as an LP. Europe has been using capital to substitute for rules it never passed: a shift from pay-as-you-go to capitalized pensions (politically very costly - ask Macron just how hard it was already to adjust retirement age), an IORP directive that caps what pension funds may do, a Solvency II that priced insurers out of unlisted equity until its 2025 review. Come to think about it, the EIF’s permanence that we blamed a few paragraphs ago is not a capital problem at all, it’s the shadow of a rules problem. Every euro the State pours into a fund-of-funds is, more or less, the cost of an ERISA equivalent it hasn’t written.

And this is why I’ve come to see Tibi - the program that got French insurers to commit billions to venture - as one of the smartest things the French State has done for our industry, and it’s not an incentive working as a rule, but as a move of pure politics. Nobody got paid to sign Tibi: no tax break, no matching check. What the State did was organize a pacte de place: it labelled the funds, coordinated the commitments, and applied the kind of pressure only Bercy and the Élysée can apply, with the DG Trésor describing its own role as “facilitator”. The label created a list of funds that passed a common filter - basically the who’s who of French Tech.

Tibi aims to replicate the 1979 prudent man rule’s effects without the hassle: the State isn’t absorbing financing risk (it brings no capital) nor regulatory risk (PACTE had opened assurance-vie to FPCI and FCPR a year earlier (5)), it’s absorbing the allocator’s reputational risk, the one that makes each insurer rationally unwilling to move alone. And it worked: fifteen insurers among the 21 investors of Tibi I, €2.3bn committed by mid-2021, €6bn over the first phase and close to €30bn with co-investments, then a third phase launched from the VivaTech stage this June that brings commitments to €13bn on a €15bn target, half of it pointed at deep tech and AI infrastructure. So Tibi is the intermediate case we were missing between the State-as-LP and the private market: the State doesn’t supply the capital, it manufactures the market in which private capital agrees to supply itself. The visible hand of the market with a very invisible (because non-existing) budget.

Instructively, though, the impact is limited - probably because of this intermediate form-factor: the IGF’s evaluation from October 2025 found where the program stalls: (a) 40% of the supplied capital to Tibi funds remains from government-backed entities (old habits die hard) and (b) not the insurers’ willingness, but their asset-liability management is liable (that’s a lot of liabilities, I know). Coordination took the private LP base as far as it could go, and what’s left in the way is … more rules.

There is a darker reading of the policymaking instrument, which we’ll develop in the next section. Rules absorb regulatory risk for whoever abides by them, but they can also be manufactured into moats by whoever sits closest to the pen.

Uber and Airbnb tackled the play on one end: operate in the grey zone, grow past the point where enforcement is politically feasible, then get the rule written around you. a16z’s Washington office runs it from the other end: write the rule first, then invest in what it allows. Whether a crypto token is considered a commodity or a security is one paragraph - but that paragraph is worth billions. I said in the demand section that political risk, unlike market risk, can be lobbied. Well, this is what it looks like when the lobbying is done at industrial scale.

Let’s recap the State side of the ledger before moving on. Three instruments, three risks: demand removes market risk, capital defers financing risk, rules absorb regulatory risk. The American ecosystem was built on the first and the third, and the European one mostly on the second. That difference is substantial, we’ll come back to it.

Before that, though, someone very familiar with strategic moves has been arguing that the (US) State is now doing all three at once, for a handful of firms, and still calling the result venture capital when it is in fact something else entirely.

The Era of Venture Mercantilism?

Sinéad O’Sullivan is an economist with an aerospace background who spent years on the strategy side of Harvard Business School (6), and her piece Honey, We Need To Talk About Venture Capital went around in the VCsphere last November. I really encourage you to read it if you haven’t.

The TL;DR is a harsh take on the industry: venture capital started as an asset class that took risk, and it has evolved into one that spends billions making sure it never has to take any. Her explanation goes like this. Early funds were small, so they could live on variance (one win of regular size was a fund returner) ; success made them larger and larger, moving from $100m funds to $10bn funds, and a $10bn fund cannot afford variance - it needs certainty (7).

So the megafunds moved upstream into policy and built four levers to manufacture that certainty: narrative (podcasts, essays, “It’s Time to Build”, American Dynamism), network (the same handful of firms moving together into the same few “consensus winners”), regulatory (a16z building a Washington operation and now openly lobbying to rewrite defense procurement rules around commercial technology) and industrial (sell to the customer who never goes bankrupt - Palantir, Anduril, SpaceX, OpenAI).

The result is something she calls venture developmentalism: the State outsources innovation to its financiers, the financiers outsource risk to the State, and the whole thing merges into a “venture-industrial complex” that increasingly resembles the integrated Chinese model where the State’s role is … basically everything. Quite the hot take for a liberal country.

Sinéad dates it in three eras - Venture Capitalism before 2008 (the craftsman times), Venture Developmentalism until 2020 (riding the ZIRP era and growing big), Venture Mercantilism since (integrated venture-industrial complex) - and lands the essay on what she calls the real fork in the industry: on one side, “outcome writers” who convert government priorities into returns ; on the other, “degenerate gamblers” still playing the 2003 game, buying lottery tickets into startup companies for 2% a year. Little Venture, in her words - Craft, in ours - is in the nostalgia business.

Ouch! You don’t really have to think for long to know that I find myself in the second camp, so while I’m reasonably okay with being called a degen and call it a day, let’s take the argument seriously and test it. Because in large part, I agree with Sinéad.

In fact, that’s what I spent the first part of this essay and a good chunk of the series on: her four levers map onto our government-backed instruments:

• Industrial alignment is the demand instrument.

• Regulatory leverage is the rules instrument, and she’s right that it has become alpha: I wrote a few paragraphs ago that political risk can be lobbied, and she’s simply describing what that lobbying looks like at industrial scale - staffed offices, draft legislation, general counsels publishing the alternative rulebook in plain sight (8).

• Network leverage is what I footnoted in Edition #3 as “Consensus Capital” (we’ll get back to that in the next edition): is Anthropic at Series E a good deal? Damn sure - everyone agrees, so everyone is in it.

• And narrative is the media factory from Edition #5, with one twist I’ll admit I hadn’t seen: I described media as a broadcast machine pointed at founders for sourcing and LPs for fundraising. Sinéad points out it’s also pointed at the political class, and that a good word duly repeated through content eventually becomes a line in a bill. She’s right about that too.

Here’s the thing though: the State was there all along.

Her first era framed as Venture Capitalism, the “risk and discovery” years of pre-2008 - … never quite existed. Go back to the demand section: the Valley’s first product - integrated circuits - had one customer and it was the federal government. The asset class she describes as pure discovery was propelled by a pension rule and a tax cut in 1978-79. Even Google grew out of federally funded research (an NSF grant specifically). Uber and Airbnb weren’t a break with venture’s origins: they were venture rediscovering that the rules instrument existed and could rule in their favor. The State didn’t capture VC somewhere around 2010. VC was born State-adjacent, somehow forgot about it during a couple of decades, and remembered when the cycle turned. We don’t have enough history to know for sure but it’s probable VC will do this every time the market’s mood swings.

Sinéad also argues that the risk hasn’t disappeared, it has “simply been nationalized”. In the language of our Edition #1, risk was indeed never destroyed: it kept travelling down the social ladder. New Bedford merchants, then adventurous heirs, then pension funds in 1979, and now - through procurement, subsidies and sovereign co-investment - the taxpayer. What she calls capture, Thought #1 called “the surface of risk-bearing expanding in profile”. It is the same observation, made with a different mood.

Overall, I think Sinéad’s main argument reconciles well with our own frameworks: her “outcome writers” are funds whose dominant risk is absorbed by the State. Their deals hinge on one uncertainty before the term sheet - which, as we said in Edition #1, is the purest VC deal there is. But that risk is not market, and it’s not technology, it’s political. Thus, it does change which pillar the game is played on: the edge isn’t in Selecting anymore, it’s in Winning at the political layer. Access to the State, the market risk absorber becomes the craft. Her “degenerate gamblers” are funds that still underwrite the dominant risk themselves: the founder, the story, the market that doesn’t exist yet. That’s the workshop from Edition #5, the last one in an industry of factories. Both are venture capital, they just live on different pillars.

And if you’re following here’s what we said about factories in last edition: they commoditize. Every fund across the US and Europe now has (or is building) a “sovereignty” vehicle and a defense-tech thesis ; “American Dynamism” has been forked into a dozen national variants. When the State becomes the absorber that everyone can point a fund at, proximity to the State stops being edge and starts being table stakes - and the alpha migrates back to whoever can still form a belief about a company that has no customer yet. Her own last prediction concedes this, in fact: she expects the frontier to move to the periphery, to small funds in the gaps the policy machine can’t reach. That’s our Thought #5.

One last thing before we leave Sinéad: she is describing a State that has become every line at once - customer, capital, and rulemaker. When one actor absorbs all the risks, you no longer have an ecosystem, you have a plan (as in the USSR kind of plan), which is precisely the case she went to study in China. It’s also, from the other side, a useful way to ask what Europe has. But before Europe, the other lines of the ledger - because there are more parties than the State to the ecosystem.

Ecosystem: the other Shock Absorbers

“Welcome to the jungle, we got fun and games”
Guns N’Roses

In 2019, I wrote a book humbly called The Jungle Book, making systematic parallels between the startup ecosystem and a jungle in a long 400-page metaphor (which shows you this series is still at moderate length in my habits).

I’ll spare you the long route: the State gets the longest section in the edition because it’s the largest line, but the whole point of an ecosystem is that it has several actors, or according to our framework, several risk-absorbers. Let us go through the others, with the same question each time: which risk does this actor take off the VC’s hands, and what happens when it’s missing?

Universities and labs absorb talent risk and, in deep tech, a good part of the technology risk itself. The Valley’s origin story is a professor (Terman) pushing his students out of the building, and it’s easy to forget how much of the industry’s raw material is produced on a public research budget before any investor shows up: we mentioned Google earlier, but most of what’s called “deep tech” in France is an SAS spun off CNRS or INRIA with a deck and funding. In the late 2010s, I used to be vocally cynical about universities’ entrepreneurship programs - they missed the point and talent was better off trained building inside startups rather than learning how to build a business plan and winning a pitch contest. Years later, I’d say the level of those programs has materially improved, perhaps because the European ecosystem itself has improved and experience has trickled down from a generation of founders and operators to the next.

Big Tech absorbs two types of risks, very distinct. Talent risk: the ex-Google, ex-Stripe, ex-Palantir pipeline is a subsidy paid by shareholders of public companies to train people who will later quit and found things. What I nuanced just earlier still stands: you learn more by doing than on university benches. And liquidity risk: an acquirer who never runs out of money is the buyer-of-last-resort that makes a VC’s downside publicly tolerable (the reason “acqui-hire” is a word). Europe has the first (Google Paris, Amazon Luxembourg, the Meta FAIR lab) and very little of the second. Unfortunately, it’s the second one that shows up in DPI. That’s a conversation for a later edition though, so I’ll leave it there - note only that the same balance sheets are now financing the frontier labs directly, so that’s financing risk too.

Angels and operators absorb the financing and talent risk before VC’s risk: the pre-seed check, the first hires … they lend credibility to founders. And this line has a property none of the others have - it recycles. Exits produce angels, angels produce founders, founders produce exits, and that loop is what makes an ecosystem compound rather than merely exist. The PayPal mafia is the canonical example ; France’s version runs through Criteo, BlaBlaCar and Alan alumni, and it’s about fifteen years younger. I’d add - and you’ll forgive the self-interest, since this is what my own firm is built on - that operators are the most underused absorber in the ledger: they take the risk that VC can’t underwrite at all, the one about whether the founder knows what they’re doing, and they take it with their time rather than their money.

Corporates absorb market risk, in theory, through pilots, proof-of-concepts and procurement. In practice they mostly defer it: a POC is a subsidy that looks like a customer (hence the wide gaps when looking at contracted ARR and deployed ARR), and “POC purgatory” is the private-sector cousin of the zombie startup - a company with twelve logos on its website and almost no real revenue. Corporate demand only removes the risk when it becomes a contract with a budget line, which is rarer than the open-innovation departments of the CAC 40 would like you to believe. Same test as the State: does the money arrive as runway, or as revenue?
On financing risk, they absorb it twice over: as LPs in funds (corporates are a real share of the European LP base, more so than in the US) and directly, as CVCs. But corporate capital is strategic capital, which means it’s allocated on proxies - fit with the parent’s roadmap, optionality on an acquisition, a slide for the board’s innovation update - the same way subsidies are allocated on TRLs. And it’s very cyclical money: CVC programs are launched when the market’s good and quickly shut in downturns, and a fund whose anchor LP is a corporate innovation budget has a fundraising problem from the first quarter the parent misses its numbers.

Public markets absorb exit risk, and they are the one absorber that has been shrinking globally, in Europe and the US too. Everything about this belongs in one of the later editions (about liquidity), so a single sentence: an ecosystem without a liquid exit market has a VC industry that can raise but not return. Thus that cannot raise again.

Now let’s put the lines next to each other. Here’s the ledger for the three ecosystems that matter, with the risk-absorber in each cell.

Read across, not down. The Valley has a different absorber on every line. Europe has the same one on two of them - the State. China has the same absorber on every line, which is what Sinéad went to study and which, in our vocabulary, isn’t an ecosystem at all: it’s a plan.

Thought #6

I opened this edition with the 2027 French election and a question about regime change, and I think the ledger helps answer it, because it tells which lines a new president can actually touch.

• Capital moves first: it’s a budget, it gets voted every year, and every incoming government can reorganize Bpifrance, rename France 2030, announce or cut a plan. In an ecosystem that partially lives on government-sponsored money, it’s the honeypot.

• Rules move, but not easily: whoever sits at Bercy in 2028 will inherit BSPCE, PACTE, the CIR and, hopefully, EU Inc. Undoing any of them costs more political capital than passing them did. Still, I am quite convinced that whoever wins in 2027, they will be high on popular approval and will be willing to spend such political capital. If such is the case, it will take time, anyway.

• Demand is the question mark. It hasn’t really been in the French playbook yet (buying from startups I mean, not from national Big Corps): it will be interesting to see whether the DGA, the hospitals and the ministries start buying from companies younger than their own procurement code, and under what political agenda. After all, we got Saudi-funded manga theme parks before we got GTA VI so anything can happen.

So there is a slight shift from the pragmatic view that was mine before I started writing this article. I can bear rules (because they move slowly) and demand (because they’re not really here anyway) not going the startup’s way. But the ecosystem still needs capital, and that would be the thing to watch - with the State being a prime actor at LP-level and at direct investment level. The president and government do have some impact.

Anyway, here’s the thought for this edition: An ecosystem is a ledger of who absorbs the risks VC doesn’t, and the State is the largest line on it - a good absorber when it removes a risk, a bad one when it defers it. America funded customers and wrote rules. Europe funded funds and hoped. That is the difference, and although an election moves how much sits on which line, it doesn’t structurally change which absorbers are there and which aren’t.

And the reason why founders flock to Silicon Valley globally isn’t that there’s some magic at work in the Mission District, it is that it’s the ecosystem where the most risks are absorbed: talent risk is low (everyone’s an operator), capital risk is low (all VCs are there), market risk is lower (more B2B customers with budget to spend), regulatory risk is honed (because a16z has a lobbying department in DC), exit risk is lower than elsewhere.

Next edition, we diffract: solo-GPs, SPVs, deal clubs, megafunds - the wrappers the industry is inventing when the standard one no longer fits the risk. Stay tuned!

PS: I started this Summer series with the intent of not delegating any thinking or writing to AI. It turns out I had much more to write about than I initially thought, and the project has started to outpace me. As such, I caved and, while I still refuse to outsource my thinking, I have started to use LLMs to turn some of my notes and bullet points into first drafts of paragraphs. I then rewrite it all in my own voice, but the underlying palimpsest may show through here and there. I’d rather be upfront about it. In the spirit of the last edition, once you know where the craft lives, you can industrialize around it.

Footnotes

(1) On that note, I recommend the excellent TV Series ‘For All Mankind’ that dives in detail on the Apollo program and what it could have become would it have received continued funding after the 1960s.

(2) Cour des comptes, Les aides publiques à l’innovation des entreprises, 2021. The same report notes that close to 40% of the CIR goes to large companies, which is a different debate…

(3) A conversation that might end soon, on account of an initiative that is not French-only. In March 2026, the Commission proposed the “28th regime“, an optional pan-European company status - incorporation in 48 hours for under €100, and harmonized employee equity taxed only at sale, the whole thing explicitly designed to give the companies that left a reason to come back. Because they do leave: by the Draghi report’s count, 40 of the 147 unicorns Europe produced between 2008 and 2021 have moved their headquarters abroad, mostly to the US. Adoption is expected around 2027. Which, for a rule, is fast. Like St Thomas, I’m still waiting to see.

(4) The instrument only moves fast when founders force it to: the Pigeons movement in 2012, when the government appointed by newly elected François Hollande had plans to tax capital gains at income tax rates and was pushed back within weeks by what was, essentially, a Facebook group. I was merely out of prépa at the time, so I only followed it from afar - but my co-founder Christophe, then a GP at ISAI, was very close to the epicenter of the movement. Arguably the founding moment of what became La French Tech a year later: Bpifrance was created by the law of 31 December 2012 (merging Oséo, CDC Entreprises and the FSI) and the French Tech label in 2013. Not all credit due to the Pigeons of course (Bpifrance was part of Hollande’s campaign promises), but not a coincidence either, I think.

(5) As recommended by Alice Zagury, Jean-Noël Barrot and to some extent yours truly in our 2017 report (proposal #9) - you’re welcome!

(6) I initially thought about naming this edition ‘Whoever wins in 2027’ ; a nod at the late Clayton Christensen, the superstar HBS strategy teacher who in 2012 wrote an op-ed ‘Whoever wins on Tuesday’ shrugging at the Obama/Romney election: the president doesn’t control the economy, the rules do. The thought of this article, as you’ll see, is similar. Sinéad O’Sullivan, on her end, works closely with Michael E. Porter, the other superstar HBS strategy teacher, and her argument’s spine is very much a Porter framework (structural positioning). Porter & Christensen had a couple of feuds under their belt about disruption, and it echoes - as you’ll see - where O’Sullivan and I differ: she describes how the incumbents mega-VC funds built a moat, I’m more interested in seeing that there are new peripheral frontiers for new entrants. Same analysis, different reads. ‘Whoever wins’ also meant Porter/Christensen. But this felt all too obscure given what this article is really about.

(7) Similarly to large Private Equity funds that won’t have 10x returns but that cannot afford to have a loser in their portfolio construction. This doesn’t yield a 3x DPI, but this is not what a fund this size is after. Only, for VCs, you can’t build a portfolio of non-losers: there’s still a large chunk of inherent risk and even if you somehow managed to, fixing every loser to a 1x - a “perfect” no loss portfolio - only moves the MOIC to +0.4x on average (or almost nothing in DPI terms).

(8) And in the terms of Edition #5, it can be a factory, either a standalone one or as part of a Platform play. Guess who’s good at lobbying? a16z!

If you missed it, you can read:

• #1 - Financing Risk : a brief history of Venture Capital

• #2 - Chance, the Wrapper: VC as a Financial Product & Asset Class

• #3 - Sum of the Part(ner)s: Why VC Partnerships Fail

• #4 - The Biggest Cliché in VC

“You keep telling me this, and telling me that (...). Here I am baby, practice what you preach”

Barry White - Practice What You Preach

I recently re-watched the excellent Moneyball, featuring Brad Pitt and Jonah Hill. The movie recounts the true story of the 2002 Oakland A’s and how their manager Billy Beane attempted to assemble a competitive team on a fraction of his competitor’s budget. Famously, the 2002 Oakland A’s won a 20-game streak following the unglamorous technique devised by Beane, which I’ll tell you about just after.

Beyond the great acting skills of the cast, the compelling storytelling of Aaron Sorkin (whom we also owe Steve Jobs and The Social Network) and the overall entertaining pace of the drama, I always thought this was a great movie about entrepreneurship - I even included it in a 2017 article I never finished called “12 movies for entrepreneurs”. Fortunately, I did write the Moneyball part, and if you’re interested, the paragraph is served in full in the Footnotes (1).

This is not the Moneyball-for-VC essay though: that essay has been written a hundred times and it always concludes “be more data-driven”, pushing VC towards a pure numbers’ game. No, what struck me on rewatch are two other things.

First, a lot of the practitioners - in any given industry - will eventually become lazy and stop challenging the frameworks, the models, the patterns, the habits of their practice. “We’ve been doing this for a long time” tells the head scout to Billy when he exposes his new strategy, thinking he knows better. And of course, with practice comes experience and ease of doing things. But in an industry like VC that is about deciphering the future, holding practices for too long might be counterproductive, or dangerous. Challenging practices in VC is the first objective of this essay.

Second, the movie reveals how efficient practice of a craft is not glamorous. It’s often boring. The only thing Beane cares about in hiring new players is statistics: “does he get on base”, thereby compressing his act of selecting at a single chokepoint. He selects aging players, players with a funny throw, players with sulfurous reputation … that get on base. Likewise, we highlighted in the previous essay that Social Capital’s bet on Groq was not one for fame and posters - nobody cared about AI inference chips in 2016. Nobody except those dedicated to the craft of VC that saw something in it and made a bet that proved ahead of its time 10 years later. The second objective of this essay is to remind myself that, however unsexy, the craft and practice of VC takes discipline and focus over hype.

Let us then dive into the practice of Venture Capital, or what we in StartupLandia call “VC Ops”.

Ars Gratia Scientiae

You might have partly recognized MGM’s motto ‘Ars Gratia Artis’, which translates to ‘art for the sake of art’. Similarly, I named this section ‘art for the sake of science’ because I wanted to briefly expand on a wording choice I made since the beginning of this exploration.

I keep referring to the act of practicing Venture Capital as a “craft”. A craft, according to the universal truthteller that is Wikipedia, is an occupation that requires particular skills and knowledge of skilled work. To me, this is what Venture Capital is all about: it’s a combination of people skills, analytical skills, communication skills.

The word matters because it eludes as much as it designates. Venture Capital, no matter how much data and statistics you can build with it, is not a science: there is no equation, no measurable way to find successful founders, no reproducible experiment over time, no control group because every fund is a small sample of unrepeatable events. But it is not an art either. Art answers to nothing but itself, the VC answers to LPs. You can’t be squandering cheques in the wild to whatever founder you deem interesting without structure and strategy. Ars gratia artis is a luxury no GP can afford.

A craft sits precisely between the two. It is mastery exercised under constraint, learned by repetition rather than theory, and transmitted the way crafts have been transmitted since millenia: by apprenticeship. There is no degree that makes you a venture capitalist. If I may grossly paraphrase a French saying, “it’s by venture-capitalizing that you become a venture capitalist”. This is why, as we saw in Edition #3, partnerships remain the industry’s only training infrastructure. The analyst-to-partner ladder is a guild structure, apprentice to journeyman to master, a type of organization that predates the industry by at least a few centuries.

Of Craft and Industry

Every guild eventually got swooped by the scale industrialization required, and likewise every craft eventually meets its factory. The last couple decades of VC - we’re nearing the 20 years of the 2008 crisis - powered by ZIRP (and subsequent AUM increase leading to more capital needing deployment) have been one long industrialization attempt: platform teams, sourcing engines, CRMs scraping Github and LinkedIn, “signals”, and at the extreme, the Tiger model of c. 2021 inspired by the most ruthless Wall Street organizations: diligence outsourced to consultants, no board seats, term sheets in 48 hours. An assembly line for capital deployment.

What’s interesting is the typology of what industrializes well, and what doesn’t. Industrialization succeeds whenever the task is objective: coverage, pipeline hygiene, process, speed. It stalls whenever the input is a person and the output is a belief, or said differently when the task is subjective. This is what I actually dislike about lots of the Moneyball-for-VC approaches: Beane did not really replace judgment with statistics. He compressed decades of scouting judgment into a single question: does he get on base. The statistical approach he built around it was only as good as the craftsman’s know-how it was supposed to encapsulate. The craft question for any fund should therefore not be “how much of our process can we automate” but “what is our version of ‘does he get on base?’ “. Put differently: what have we learned to decipher systematically by doing the craft can we confidently encode into a program or automation? (2)

Another thing to note about the movie while I’m at it: metrics in sports eventually won so completely that it stopped being an edge: within a few years, every club had an analytics department, and the competitive advantage migrated elsewhere - requiring more craft to find new edges. When a craft is fully industrialized, its alpha evaporates and it becomes an attrition war. Hold that thought until the end of this essay (spoiler alert: we’ll be talking about AI in VC).

Art, Science, and the Lifecycle

Another interesting thing is that the mix of art and science in the VC craft is not fixed. It moves through the lifecycle of the company in question. At the early stage, a company is its founder: it has no product, no metrics, sometimes no deck. Underwriting at this stage is close to divination, the “fortune telling” of Edition #4. Investing in it requires flair (both in the English and French meanings - the latter being close to intuition or having a knack for something). A few years later the same company has cohorts, KPIs, a board, governance: it has collected data on record, and data can be crunched and analyzed. The craft shifts from art toward science as the object matures.

Which yields a corollary that explains a decade of market structure: the more scientific the stage, the more industrializable it is ; and the more industrializable, the more competitive. This is why crossover and quant capital invaded growth first: growth rounds are underwritable from a spreadsheet business plan. The early stage, in my view, is where the craft remains irreducible, precisely because it is non-measurable. It remains the last workshop in an industry of factories. For now.

The Four Pillars of the Craft

If you break down the most basic components of a VC’s job, there are four recurring activities in the operations: finding companies to invest in, picking companies to invest in, managing to be accepted by the companies you pick ; and then once you’ve invested, supporting them in the best way you can.

I do not intend to offer a handbook on how to perform each of those activities. Rather, (i) with respect to what we said in the intro, challenging the way it’s usually done (if there is an usual way) ; (ii) seeing where it gets you when you push the cursor to the maximum for this specific task and (iii) deciphering for each what part of this activity is craft, and what part can be industrialized?

Sourcing: “Fantastic Beasts and Where to Find Them”

Sourcing is often confused with finding companies. However, the point of sourcing is not coverage - coverage industrializes easily (3). The point - ideally - is primary access: seeing founders others do not yet see, or being one of the people they will take a call from before the market arrives. The best sourcing engines therefore tend to begin with an unfair access point: a community, geography, operating network or domain - and compound from there.

Note that I (willingly) did not use the word proprietary. Proprietary access implies exclusivity. Primary access implies timing and position. Great founders rarely stay proprietary for long: they know how to navigate fundraising, and good investors eventually find good companies. What matters more is primary access: seeing the company early enough to form your own view before the market has formed one for you, and getting bonus points with the founder because you can rightfully claim you believed early, made intros and pushed them.

That distinction matters even more depending on your strategy. A lead needs enough primary access to build conviction, win the allocation and help assemble the round. A follower can arrive later and still make an excellent investment, but by definition more of the information has already been processed by the market - and often priced accordingly. In both cases, the deal is seldom proprietary: the convinced lead will help assemble a roster of angels and follower funds that add value ; the convinced follower will help find a lead. Yes, it also happens that a lead fund wants to take the whole round because they formed a unique conviction and want to increase ownership, and thus require exclusivity. It is an increasingly demanding thing to ask founders though.

Anyway, the name of the game here is to build a recurring engine for primary access. And I think the useful way to think about such an engine is to stop treating “the startup” as the unit of sourcing and to look one or several steps upstream. The question thus becomes: who repeatedly encounters the founders you want to meet before you do? Depending on stage and strategy, that can be angels, earlier-stage funds, portfolio founders, operators, university labs, accelerators, domain experts or LPs.

Uh,uh - so you’re telling me: ‘use your upstream network to source’? Is this Venture 101? This is indeed basic, but saying it out loud also clarifies what automation can and cannot do. Most modern “sourcing engines” are in fact discovery engines: LinkedIn scrapers, databases, GitHub signals, job-change alerts, incorporation data, LLM scoring. They are excellent at telling you who exists, which is why they industrialize so easily and eventually become table stakes. Discovery tells you the company exists and can be automated ; the access engine makes you one of the first people the founder wants to speak to, and it cannot.

And one of the best ever at this game of unfair advantage was Chris Sacca, whom we already mentioned in Edition #2. His Lowercase Capital I fund invested in Twitter, Instagram, Uber, Stripe,Twilio, Kickstarter, Docker. Talk about primary access … How he did it is the more interesting question. And the answer is dead simple: Sacca had just inserted himself into the social graph from which companies were being born. Google (where he worked on Special Projects) gave him the first circle. Evan Williams, whom he knew there, later asked him directly whether he wanted to invest in the Odeo side-project Twttr (tka Twitter, nka X). Twitter then put Sacca inside an unusually dense network of founders, angels and future founders. His unusual location in Truckee, CA and original style reinforced it further, with people like Travis Kalanick spending time with him while Garrett Camp and Kalanick were developing what became Uber. The early Uber financing was devised at Sacca’s own house!

And then the mechanism compounded: meet a great founder, become useful to them, meet the talented people around them, invest in some, become one of the first calls when the next company appears. From Google to Odeo/Twitter, to befriending angels and YC backers to more founders to those founders’ friends. Nothing was particularly proprietary about it: the good companies still circulated. What Sacca built was recurring primary access, repeatedly placing himself close enough to companies that he could form a view on before the market did. Granted, at a time when “seed” round wasn’t a thing, when he could scoop up valuations that one could only dream of today and when Silicon Valley was about to explode with SaaS and apps opportunities.

By the time Evan Spiegel approached him directly about Snapchat, the machine had compounded so far that founders were coming to him. The craft of sourcing bore fruits.

Sacca passed on Snapchat though, a useful reminder that sourcing only gets you into the room and selection remains a different craft. Luckily, this is our next section.

Selecting: “Do you believe in this thing or not?”

Selection is interesting because it’s the pillar with the most mythology and the least verifiability in the industry. It’s also the industry’s favorite self-image because it strokes the brainy nerve that led many investors to doing this job in the first place. Ask any GP what they do and they will tell you they are pickers. This comes packaged in a oh-so-common way of presenting it: not necessarily ‘we do more due diligence than others’ (although this comes out occasionally), but often with more magic and intuition about it: ‘we excel at pattern recognition’, ‘we can tell founder quality’, ‘we time the market’ … grand gestures suggesting an uninspectable skill whose mystique is conveniently impossible to disentangle from the skill itself (4). This usually pairs well with disregarding investors who ‘pray and spray’, a bit like a local pub champion would disregard a blindfolded monkey throwing darts: “no no no, we don’t do that here”. However, in order to hide the mystique part of it, the usual approach is to treat picking as a teachable framework, and thus a trainable skill.

Here is the problem: selection is the one part of the craft you cannot really practice. Billy Beane could measure “does he get on base” every single game: weekly feedback made his chokepoint trainable. A VC’s picking quality is revealed a decade later, mixed up with price, access, luck and the market’s mood (Edition #2 has entered the chat). Over a career, you get perhaps three or four (and max seven, if you’re Arthur Rock) fund cycles of resolved feedback, a bit more if you invest in parallel funds, but that’s still a few dozen of meaningful data points, most of which arrive too late to change how you invest. There is no deliberate practice at a ten-year feedback loop, there is only retrospective anointment or dismissal.

And because the real feedback (DPI) is unbearably slow, the industry built itself a proxy: the hit rate (of which there are many definitions, but let’s say it’s the % of your investments that end up among the industry’s best outcomes). Let’s look at hitrate data.

DST Global’s Cole Rotman assembled a list of the Series A lead investor in every $5bn+ venture-backed company (exc. crypto, bio, energy and China) from 2012 to 2024. The data reveals huge concentration: 9 firms - a16z, Sequoia, Accel, Benchmark, Index, Lightspeed, Matrix, Khosla and Greylock (5) - led more than half of all such Series A rounds. But if you divide each firm’s winners by all the Series As it led, the picture inverts: a16z “only” converts roughly 2% of its ~450 led rounds while Benchmark converts roughly 10% of its 63. Rotman notes that between 2013 and 2018, about 30% of the Series As Benchmark led became $5bn+ companies, or nearly one in three. Across the top firms, the average conversion is about 3% (it drops to 2% if you remove Benchmark). At Series B, Benchmark converts 19% of the rounds it leads, against 13% for the next best (Y Combinator). Stanford’s Strebulaev (whom we already trusted in Edition #2) has similar findings, but since 2020: a16z produces the most unicorns in absolute terms (56) but from over 1,000 investments, a ~5.5% rate - and his top performer is Firebolt Ventures with a 10.1% unicorn rate (13 unicorns out of 129 investments since 2020) (6).

Now before anointing anyone (not that it’s in my prerogatives), we should do the larger math. 11 outcomes from 454 shots versus 6 from 63: a16z bought roughly twice the outcomes at seven times the volume. Okay, but that doesn’t tell us much. Which machine is “better” depends entirely on the fund behind it: six $5bn+ outcomes at Benchmark ownership returns a $500M fund several times over, while a16z’s AUM base requires absolute count (56 unicorns since 2020), not conversion. You can’t judge those on the pure merits of ‘who’s more skilled’, those are two different strategies at play: betting on the power law (picking better) versus indexing it (accessing better).

And interestingly, this is exactly what we said earlier about craft and industry. Benchmark practices selection as craft: concentrated, artisanal, conviction-priced. a16z industrialized selection into coverage. I’d wager that neither firm is confused about which game it is playing.

Which brings us to what you can’t really see of the craft from the data, but only intuit: 63 led Series As in 13 years is ~5 per year, and for an entire partnership like Benchmark it means on average one deal per partner per year. For someone who likely sees 1,000+ founders. The invisible part of the selection craft is refusal. Conviction only works on a budget, and the more concentrated your fund is, the more carefully you have to spend conviction.

So what industrializes here, and what doesn’t? Rejection is easily automated: screening, scoring, running against an industry “funding napkin” of expected growth or revenue figures, timing - it’s always very easy to find a reason to say no (especially if you’re using preposterous benchmarks, and increasingly as the company moves to the next stages). What requires craft is the other decision: the yes formed on a company that is still just a founder and a story, against consensus, with no data to back it. That is not analysis, it is belief formation.

Hence a take that will certainly raise a few eyebrows in my pictured wide readership of engaged investors: investors should not try to make selection their edge. Not because it isn’t one, Benchmark’s numbers prove it is the most valuable edge in the business ; but because it is the one edge that cannot be built on purpose. Its feedback loop is too slow to train against, its proxy metrics are paper, and its masters are identified only in the past tense. Selection edge is not engineered, it is revealed. Build the edge where the feedback is fast enough to compound, and if, ten years later the market (that is, Ilya Strebulaev, Cole Rotman or anyone nerd enough to crunch the data) informs you that you were a great picker all along, accept the compliment and move on.

Of course, conviction is worthless if you cannot get into the round. Benchmark’s hit rate, a16z’s absolute number of unicorns are also a winning story: they didn’t just pick those companies, they were chosen back. Which is our next section.

Winning: “All I do is win, win, win no matter what”

When I first entered this business, I carried a mental model of supply and demand: there is less capital than there are companies being created, therefore investors hold the bargaining power. Founders pitch, investors dispose. I’m sure I’m not the only one who thought that way, and it takes about one competitive deal to unlearn this. The reality is that capital does not spread across the market, it herds eugenistically: dollar organisms overwhelmingly want to swim where other dollar organisms are already going. A “hot round” is nothing more than consensus made reality. So the actual market structure is an inversion of the rookie model: infinite capital is available to the same few companies (7), and for those companies, the founder disposes, and the VC must win.

And here we hit a wall that I want to point at rather than hide: there is no data on winning. We just spent a section dissecting in hit rates under every angle (conversion tables, Stanford rankings, Google Sheets lovingly maintained by growth investors, etc.). There’s no equivalent for win rates: the percentage of term sheets issued that got signed. Nothing, besides a few partners here and there hinting at numbers that suggest 85-90% winrate is top-tier, as one could imagine. And that’s not because firms don’t know it - every fund obviously tracks its lost deals with the bitterness of a jilted lover - but because nobody reports it. It’s easily understandable, too: the hit rate is flattery, the win rate is confession. Publishing “we picked Uber” is marketing ; publishing “we offered and were turned down forty times” is an autopsy. Even those brave firms that show their anti-portfolio (the deals they passed on that became great) don’t share their lost portfolio. Winning is both the most operationally obsessed-over and the least publicly documented pillar of the four.

What distinguishes winning from the other pillars as well is that it’s the only pillar where the evaluation reverses. Sourcing, selecting, supporting: you act on the company. Winning: the company acts on you. The founder is running their process on you ; and the good ones run it seriously. They will back-channel your portfolio, call the founders you disappointed as much as the ones highlighted in your website, ask how you behaved in the down round, in the failure, etc. In every other pillar you are the investor ; in this one, you are the product, and due diligence is being performed on you.

So how do you win?

You can pay. Pricing up is the industrial lever: it requires no craft, only capital, which is precisely why it commoditizes instantly. Anyone with a bigger fund can outbid you tomorrow. It is also the megafund’s natural weapon, and rationally so: when your model is indexing the power law, overpaying for entry into the index is the cost of doing business. It’s a fee, not a fault. For everyone else (as a lead investor - important distinction), paying up is more akin to renting the win rather than owning it, because of how it degrades your fund economics.
And pushed to the maximum, this lever produces the most instructive specimen of the last cycle: the Tiger model of 2021 won essentially everything it wanted - highest price, fastest close, no board seat asked - and it won its way into a graveyard vintage (which might very well turn and recover, I have no crystal ball. But if you’d ask me to place a bet on it on Polymarket, I wouldn’t). I remember from my Econ 101 class at HEC that there’s a name for this: the winner’s curse. In an auction over an asset of uncertain value, the winner is, by construction, the bidder who most overestimated it (8). Which reveals something structural: winning is the only pillar where pushing the cursor to the maximum by industrializing is self-defeating. A 100% win rate doesn’t mean you’re excellent, it means you’re the market with a premium. Sourcing can’t be too good, selection can’t be too disciplined, but winning at all costs is not winning at all in the end on the metric that matters (say it again with me fellas: “DPI”).

You can be wanted. If paying up is the industrial lever for winning, reputation is the craft lever. Reputation can be decomposed into three layers that compound on different timelines:

• The intuitu personae. Because founders don’t actually choose firms, they choose the human who will sit in their board room for ten years and who they will text at 2am on a Sunday with anxious thoughts. This layer is pure craft, unscalable, and it’s why the same firm wins with one partner in the room and loses with another. It’s also, to me, the beauty of our job.

• The brand. Firm-level, decades in the making, the reason a Sequoia or Accel term sheet can accepted at a discount to the highest bid (9).

• The public record. The layer you don’t control anymore: every founder you’ve ever worked with is a reference. Horror stories travel at the speed of WhatsApp (or LinkedIn posts if you’ve outdone yourself, and the worst is podcasts - the Kalanick-Gurley relationship sure was a tense one (10)) and a pulled term sheet will be remembered for a decade. And the worst thing about reputation is that it is built in your worst moments, not your best. It’s the one asset to protect at all costs as an investor, and the one way to do it is by behaving decently with founders, always.

You can be useful. The third lever is your Firm’s OS (or talent as angel): whether you bring something genuine to this deal’s table, and whether you can credibly promise to deliver it. This is where sector focus compounds into a winning machine: the specialist speaks the founder’s language in the first meeting, has the customer intros ready in the second, and - crucially - can reach genuine conviction faster and hold it deeper, because they’ve already done the market work. And speed is the most underrated winning weapon there is, because the highest form of the craft is not winning the auction. It is making the auction not happen: arriving with primary access (the Sourcing engine paying dividends), converting conviction into a term sheet before the round is hot, and pre-empting consensus instead of outbidding it. The best firms win without pricing up not because founders are charitable, but because by the time everyone else showed up, there was nothing left to win.

Which leaves one question hanging: where does that reputation - the one founders back-channel, the record that wins you the next deal - actually get manufactured? In the least glamorous pillar of the four. Supporting is next.

Supporting: “How Can’t I Be Helpful?”

Every founder knows the sentence. It arrives by email a few days after the wire, or maybe at the end of an insipid 30-min catch-up call that revealed profound misunderstanding of what the business actually does, either way radiating goodwill and asking for nothing except, somehow, a task: “How can I/we be helpful?” Supporting is the pillar with the widest gap between what is pitched and what is delivered.

Before dissecting that gap, I’d like to take a moment to describe what I believe the base case for support should be, because I use a weird term for it, but I mean the word as a compliment: Cheerleading Capital. The cheerleaders are one the sidelines and root for their team, no matter what. Likewise, the cheerleader-investor’s contract with the founder is modest and truthful: money that closes fast and behaves predictably, no harm done, genuine enthusiasm, and availability when called upon. Do no harm is not a low bar, by the way: an investor who doesn’t consume the founder’s time, doesn’t panic in board meetings and doesn’t meddle in operations is, by founders’ revealed preference, already above the median of their cap-table. The cheerleader never pretends to be an operator or a strategist and cheers sincerely from the stands. I’d be willing to bet that most founders, pressed honestly, would rank this above half the “value-add” they have received.

It was important to me to describe the positive version, because the problem of this pillar isn’t Cheerleading Capital. It’s what we could oppositely call Misleading Capital: firms that pitch operator-grade involvement during the round (the network, the playbooks, the 2am availability) and then deliver less than a good cheerleader would. That’s the gap between pitch and delivery.
But come to think about it, I believe that the gap is structural, not moral. Two mechanisms produce it. First, support time allocates inversely to returns: the companies that consume the most help are the struggling ones, which in a power-law portfolio are precisely the ones that won’t move DPI however heroically you assist them ; while the winners, the only companies your fund math actually relies on, are winners in large part because they don’t need you whatsoever. Second, support has no end-game metrics: “we made Y intros” can very well justify for a hire or two, or maybe some revenue. But what does that have to do with the shape the company eventually takes? Granted, one thing all VCs do is intro to later round investors (11). But that’s also because it bumps their own TVPI and eases raising the next fund (Adam Smith, medudes). Therefore, every time an investor says “we helped them scale”, it is an unverifiable claim and, sold aggressively enough, it becomes the myth of a support engine that may not really exist to begin with.

And yet this pillar sends us back to the epigraph of the essay, chosen for this exact spot: here I am baby, practice what you preach. Winning and Supporting are the same pillar observed at two moments in time. During the round, you preached: the conviction, the help, the text answered on a Sunday. Supporting is where the preaching comes due and where the founder keeps score. So what does practicing look like, above the cheerleader’s floor?

To me, the craft of supporting is not about the volume of assistance. It is presence at the moments of truth. The down round, the co-founder feud, the not-so-good M&A offer, the bridge that decides whether there is still a company in two quarters. In those moments, cheering is no longer enough: the founder doesn’t need applause, they need the person who will tell them what nobody else will, and stay in the room after saying it. Every great company nearly died at least once. The support that matters is disproportionately concentrated in those few near-death weeks, and the craft is knowing which weeks those are and being findable during them. Going above and beyond, once or twice, when the company needs it.

If you think hard about it, there’s a paradox in this pillar’s reputation: it’s both overrated and underrated at the same time. If you judge it inside the current fund, the cynics are probably right: most support barely moves outcomes, and the honest cheerleader captures most of the value at a fraction of the pretense. But support’s true return doesn’t materialize in this fund (12). It lands in the next one, as winrate. Supporting, true supporting as we just described I mean, is deferred winning. The firms that hone it aren’t buying this portfolio’s performance; they’re banking trust at the moment it is cheapest to earn and impossible to fake.

At this point, if you know who I am, you should be screaming at me: ‘Younes, you run a freaking operator-investor fund. Support is how you win deals. Support is your value proposition to founders. Are you literally destroying your own firm’s core pillar?’

Not quite, because plenty industrializes in the support pillar. Notably, the moment support stops being something a partner does and becomes something a department / a platform does - recruiting pipelines, customer introduction engines, marketing muscle, content - it changes nature. You are no longer leveraging a craft, you are bolting a factory onto the fund. Some firms (not mine … well, not yet) pushed that cursor so far that the factory became the strategy. That experiment has a name with 16 letters inbetween the first and last, a headcount in the hundreds, and its example deserves to move to the next section of the essay, which precisely focuses on what some firms do as value-add operations on top of the four pillars.

VC at Industrial Scale

If you’ve followed this far, (a) kudos - maybe get a glass of water or stand up for ten seconds or something, (b) the four pillars are the craft - meaning what every fund does, well or poorly, with the hands it has. But some firms have decided to build something on top of the pillars: real operations, with headcount and budgets, designed to give the pillars industrial leverage. I’m not talking about more AUM and larger funds, although these approaches are the apanage of large funds, because they’re usually costly (13). I’m talking about the factories of Venture Capital, and I think it’s worth noticing what each factory is actually a factory of. Platform is Supporting (and, as we’ll see, Winning) pushed to the industrial cursor. Media is Sourcing and Winning industrialized into an attention machine. And the third factory (AI), the one we’ll close on, is the attempt to industrialize the last pillar everyone assumed was un-industrializable: Selecting.

Platform

In 2009, Marc Andreessen (founder of Netscape and inventor of the web browser) and Ben Horowitz launched their firm with a $300 million first fund and a heretical approach to VC: the venture firm should be built like a talent agency, not a council of wise elders. The explicit model was Michael Ovitz’s CAA - the Hollywood agency that won stars by surrounding them with services no rival could match - and the target was the incumbent model of the era, i.e the gray-haired board member dispensing wisdom and, too often, ending up replacing the founder with a “professional CEO”. This was a timely move, as Zuckerberg had just crushed everything investors knew about growth and about retaining board control. The a16z bet was therefore a welcome inversion: founder-CEOs would run their companies, and what they needed wasn’t adult supervision but “scaffolding” (didn’t know that word before Claude pushed it into every single one of its replies in the past year but I have to admit it’s useful): recruiting, marketing, sales introductions, policy. Two teams working for the founder: investors who’ve built before, and operators who clear the path. A new model derived from current market practice, or in our own words: a craft insight, industrialized.

And what I find fabulous about their vision is that the insight that founders will pick the firm that makes them more powerful rather than the one that supervises them is a Winning insight. To me, building a Platform was never meant to be Support product, it was a Winning weapon dressed as a Support product. In 2009 a16z had no track record, no brand and no public record: the three reputation layers of the Winning section were empty, beyond Andreessen & Horowitz’s own founder/angel aura - that was useful but couldn’t by itself compete with established firms. So they manufactured a fourth lever (teenagers would say “they went aura farming”, or “auramaxxing”) from scratch: be structurally, visibly, contractually useful. It was Firm OS and it worked on a scale they probably did not initially intend: from that $300 million debut to over $100 billion under management (around 18% of all US venture dollars as of 2024). The vision materialized.

It’s also worth running the math, because it explains how this whole operation stands. Platform is a fixed cost - hundreds of salaries that exist whether or not any portfolio company actually needs it - and fixed costs of that size can only be carried by management fees, which means AUM. This is a structural, dual relationship: the platform needs the AUM as much as the AUM needs the platform. Each new vehicle funds the factory, and the factory justifies the next, larger fund (that, and being good at indexing the power law). It is not a coincidence that the biggest platform in venture also belongs to one of the most prolific fundraiser in venture: the two are the same flywheel viewed from different sides (and this flywheel slowly leads to a novel type of asset management firm (14)).

Does platform even work? Let’s split the question: Does platform improve portfolio outcomes? We can’t really know: there’s no cap-table A/B testing that can tell you whether or not the odds changed materially through support. What we can point out is that the Selecting data showed a16z’s model working on coverage, not conversion. Does platform improve winrate? Demonstrably, for a more than a decade now: founders signed a16z term sheets because of the platform promise, competitors lost deals to it, and the proof is many eventually copied it. But this makes sense: platform is manufactured Winning, and on that metric it may be the single most successful operational innovation in the history of the asset class.

Which is precisely why it stopped being an edge. Through the 2010s, the platform angle entered every fund deck, every VC website, every pitch investors gave to founders - to the point where it became undifferentiated and hard to know for founders which investors actually had one and which were just pretending. But when the cycle turned in 2022, a lot of the same funds that had built platform teams to look like a16z cut them first - a silent admission that, for most firms, platform was a costume rather than a strategy. Because let’s call it what it is: a sub-scale platform is the worst of all positions - the fixed cost without the differentiation. The edge survives only where it’s defended by scale economics: a16z’s own site now makes it explicit, claiming to be the only firm with the scale to deliver on the model. The initial insight of Marc Andreessen and Ben Horowitz commoditized from craft judgment (2009) to industry factory (2010s) to table stakes (2020s): the full edge lifecycle.

So what should a fund that is not a16z take from this? Not the headcount. The lesson is the one from the beginning of this essay: a16z didn’t win because they had platform teams, they won because they had a question (‘what would make a founder choose us over Sequoia?’) and built a factory around their answer at a moment when the answer was non-market consensus. Copying the factory is not edge building. For a small fund, the honest choices are scarce: the cheerleader’s truthful contract, or a narrow, genuinely deliverable Firm OS where sector focus makes the promise credible. Outside of those two paths, my partners and I are experimenting building platform scale without platform cost using upside alignment and carried interest. We might be onto something, we might not. Time will tell.

One last observation before we move to the next factory. Even at the industrial maximum, the craft residue doesn’t disappear: no platform team attends the co-founder feud, and the moment of truth still ends up on a partner’s phone on a Sunday. The factory extends the firm’s hands but it cannot replace its presence. a16z understood something adjacent, though: that if services win the deal, attention wins it earlier. Which is why the same firm also went and built a media operation. But on that terrain, they were (arguably) beaten to the archetype by a 19-year-old Brit in his bedroom with a microphone, some nerve and a vision.

Media

In January 2015, a nineteen-year-old law-school dropout recorded a podcast interview from his bedroom in London with a $50 microphone, no track record, no capital, and by his own account, not a single contact in the venture industry. The show was called The Twenty Minute VC, the constraint in the title was the pitch (‘I only need 20 minutes of your time’), and the first guest was Guy Kawasaki, whom responded after a short, four-line cold email. Ten years later, Harry Stebbings closed a $400 million fund in four months, backed by MIT’s endowment and Rothschild’s RIT, on top of roughly a decade of episodes and north of a hundred million downloads. When asked how, his answer to the FT was one that earned it its place in this edition about edge building in VC: “We leverage media to be the best investor.”

And in this edition’s very own framework, we could say: Media is Sourcing and Winning industrialized into a single machine. Remember the Sacca mechanism: insert yourself into the social graph where companies are born, become useful, become one of the first calls. Sacca did it physically - one dinner, one founder, one house in Truckee at a time. Stebbings built a broadcast tower that performs the same insertion in parallel, into every founder’s earbuds, daily. By the time a founder pitches 20VC, they have often spent dozens of hours with Harry: they know his questions, his taste, his enthusiasm, his lost deals. The trust that intuitu personae usually builds over years of board meetings, the show can manufacture in advance, at scale, with people he has never met. It is primary access built as a media beacon. And the same broadcast engine pointed at LPs raised Fund III in a single quarter.

We could call it an overnight success with a decade of hard work on public record. But even this framing understates the craft layer beneath the factory, and this matters for anyone tempted to copy the model. The podcast alone never made Stebbings a venture capitalist: it made him famous among venture capitalists. The investing itself he learned the way this edition says the craft is always learned: by apprenticeship. A stint at Atomico, then Stride.VC, the seed firm he co-founded in 2018 with Fred Destin - a former Accel partner two decades his senior, who publicly legitimized “the podcast kid” when the industry was still smirking, and beside whom Stebbings ran a real fund: LP relationships, fund operations, leading deals. When he left Stride in 2021 to go all-in on 20VC, the attention machine finally had a craftsman running it. The guild produced him, the media just made him access it, and eventually made sure everyone knew it. Same lesson as the platform: the machine encodes a craft, it doesn’t replace it.

And if you don’t believe the craft is necessary to build an investment operation from a successful media, look at what happened in the opposite direction. In April, OpenAI acquired TBPN, the daily tech talk show that had become Silicon Valley’s favorite go-to for a reported low hundreds of millions. 11 employees, about 70k viewers a day, on pace for ~$30m of revenue that OpenAI promptly shut off, because the revenue wasn’t the point: the ads were killed. The asset being bought was the narrative channel itself, or more precisely the narrative talent of TBPN’s founders (who remain with full editorial freedom - OpenAI is not naive enough to try and buy itself a mega ad and run a psyop campaign onto the current best tech show, which is bound to fail miserably). Which means that two exit options exist for a tech attention machine: sell the attention to a company that needs the narrative vibe, or do what Stebbings did: refuse to sell it once, and instead compound it into AUM, where the same attention gets monetized every fund cycle. TBPN cashed out as a communications department. 20VC converted into a balance sheet. Same asset, opposite financial instruments. And as an aside, the fact that the world’s (#2) most valuable startup paid nine figures for a talk show tells you what attention is now worth in this industry.

So, our three questions. (i) What does the maximum cursor look like? You just read it: a firm where the media is the sourcing engine, the winning engine, the LP engine and half the support promise (the 20Sales/20Product/20Growth sub-funds are staffed by operators that the show recruited. It is a platform, but one assembled by microphone). (ii) Does it commoditize? It’s at least trying to: every fund now has a podcast, a newsletter and a content lead, and the VC content arms race is well into its table-stakes phase - even a16z, true to form, industrialized its media in-house (quite compellingly so, I have to say). But attention compounds like brand, not like software: the decade head start, the daily cadence and the back catalog are scale economics with another name, and you cannot blitzscale (15) your way to 1,400 episodes. And (iii) what’s the craft residue that resists industrialization? The host. A platform survives its founders ; a media-fund is its founder(s). Every mechanism in this section routes through one irreplaceable human larynx, which makes media the most powerful factory in this essay and also the only one with the biggest single point of failure, because it’s fully individual (16).

And we’re now down to a single industrialization attempt left. Media industrialized attention, platform industrialized service. Both left Selection untouched: the yes formed on a founder and a story. We said the Selection pillar could not be practiced, only revealed 10 years later. But someone with a very large fortune, a knack for going after difficult problems and a very quantitative disposition looked at that claim and decided to treat it as a challenge. And this deserves an entire section.

AI in VC

You might read the title of this section and think to yourself: ‘ah yes, finally the playbook that’s going to let met build edge in investing with nothing but a Claude Pro subscription’. Spoiler alert: that’s not it. So let’s get this out of the way and shrug off the boring version of AI in VC first.

Yes, every fund now runs an AI playbook: scoring engines on inbound decks, LLMs summarizing portfolio updates and drafting memos, due diligence delegated to Fable 5 and GPT 5.6 Sol, agents surfacing and executing intros from the network graph. Two years ago this was a still worth making a slide in your deck - today it’s plumbing and the value-add is as interesting to an LP as you saying ‘I use Excel and Notion’ (note: I’m not making this up, I got this directly from an LP). Additionally, you’re familiar enough with the lifecycle by now to write the entire paragraph that I won’t yourself: what industrializes easily commoditizes instantly, and nothing in the history of this industry has industrialized faster than AI tooling. You can choose to build it, or to buy it (e.g my buddies at Railblocks are doing this for the world’s best funds) but you cannot differentiate with it. Promoting your AI tooling is like saying you work with Internet: cool story bro, now tell me what you actually do with it that’s game-changing in your investment craft. And at this game, a Russian-born banker is winning hands down.

The Storonsky Experiment

The interesting version of AI in VC is indeed the maximal one, and I wrote about it at length last year, so I’ll recap and do the necessary update.

QuantumLight is Nik Storonsky’s side gig. The Revolut founder has a severe diagnosis of our industry: herd mentality, convictions that evaporate with market turns, human bias as the weakest link. His answer - as he is used to - is product: Aleph, a proprietary model fed the entire universe of venture-backed companies since the ‘90s - 10 billion data points across 700k+ companies - screening thousands of firms daily and issuing investment recommendations that the fund follows with essentially no exceptions (after meeting the founder once or twice, still). Every investment in the portfolio was machine-recommended. The team is made of engineers and quant traders, not investors. If you want more details on how this works, they’re in the article. The interesting thing is that the implicit ambition, to me, is akin to Renaissance Technologies transposed to private markets. If the Renaissance name doesn’t make you sit up, its flagship Medallion fund’s 39% net annual returns over three decades should. This is not a fund using AI. This is the attempt to industrialize the last pillar: Selecting, the one this essay declared was impossible to practice. Storonsky read the mystique footnote and was like ‘hold my beer, I got this’.

And here’s what happened since I wrote that piece and just as I was about to start writing the one you’re reading currently: QuantumLight is working, at least at something. This August, QuantumLight announced that it closed an oversubscribed $500 million second fund in six months of fundraising, which is double the debut fund of barely a year earlier. The first fund in question now has twenty-seven companies of which five paper unicorns (Together AI, Function Health, Robin AI, Fuse Energy among the names) (17). Whatever you think of the thesis, the market just bought twice as much in just a year. LP demand for AI VC is here.

But what exactly is working? Let us be cautious here, because QuantumLight’s own CEO says it plainly: the fundraising success is “built on Storonsky’s reputation and Aleph’s promise”. Reputation and promise, or in the vocabulary of this article: what the experiment has demonstrably industrialized so far is not Selecting - it’s Winning at the LP layer. A non-consensus story, a founder with an aura, a machine with a name and ambition that sound like it’s going to flip VC upside down: that’s a narrative edge, and narrative edges raise first funds. Whether Aleph can pick efficiently will be revealed the only way selection is ever revealed: a decade from now, in DPI, by retrospective analysis. The rule we established for humans also applies to machines, thankfully.

In the meantime, it’s important to highlight what the experiment concedes through its own fund strategy. Aleph invests at Series B and C. Not pre-seed, not seed. The stages where a company is still just a founder and a story are (so far) the stages the machine declines to enter.
This is exactly what we said in our lifecycle section, but implemented as an investment mandate: the factory built itself exactly at the point where companies become legible, where art has already given way to science, where the data the model feeds on begins to exist. The most radical attempt ever made to industrialize venture judgment has, by its own architecture, drawn the frontier of the craft in the same place we did by first principles thinking. The VC guild’s last workshop - early stage - remains unraided. For now (18).

There are also honest reasons to keep the champagne corked (are there really?), and they’re the Selecting section’s arguments seen under quant light. Medallion works on liquid markets: millions of trades with instant feedback and signals that repeat over time. Training models is easy because of data availability and feedback loops (not only backtesting). Venture offers the opposite regime: tiny number of transactions with non-durable rules that decay (e.g the patterns of the SaaS decade may teach exactly the wrong lessons for the AI one), and a feedback loop so slow that Aleph’s “self-healing” must necessarily retrain against markups, not outcomes: the model is currently training on paper truth, the same proxy metric we spent the Selecting section distrusting (DPI > hitrate). None of this means the experiment fails (quite the contrary, I wish lots of luck to QuantumLight in, if not removing the human out of the investing equation, at least holding investors to the highest possible data-driven standards and eventually upgrading the entire asset class), but it means the experiment cannot yet have succeeded and that everyone involved, including its CEO, knows it.

So my verdict here depends once again on the timeline. At Series C and beyond, I think the machine has got something real: those companies are spreadsheet-legible, the comparables are dense, and frankly a disciplined model may already out-select a distracted growth partner pattern-matching from memory, or even a data-driven one that doesn’t have the state-of-the-art models. A quant eventually beats even the best trader at some strategies because the data is not humanly manageable.
At pre-seed and seed, I remain where I was: the yes formed on a founder and a story is belief formation, not signal processing, and no training set currently exists for a person who hasn’t done anything yet. My weird brain thus convoked both Boolean algebra and Peter Thiel to write the following: you can data-drive your way to investing after you reached one, but from zero to one there remains an irreductible piece of belief formation about the human. Still, that does not mean that data will never tackle it: I expect the frontier to creep steadily earlier as every cohort of founders becomes more measurable than the last (their energy level read directly from their Strava, the AI reading tension during the meeting, who spoke and who didn’t and treats it as signals). The workshop isn’t safe forever - it’s safe where the data isn’t. And thankfully we won’t have data from early founder signals to DPI before another decade or two.

Let’s close on this: whatever the outcome of the experiment, notice that Storonsky is - perhaps involuntarily - the best-behaved student of this entire essay. He looked at the four pillars, picked exactly one bottleneck (the hardest, most mystique, least decipherable one), formulated a non-consensus answer, and committed to building a factory to industrialize it. No platform, no podcast, no cheerleading: one question, all-in. Which is, as it happens, precisely the final thought.

Thought #5: One Question, All In

This was a long article, but also my favorite to write and think about so far, justifiably. Because look back at the gallery we just walked through. Sacca never built a platform: he built a social graph and let it compound. Benchmark runs no media empire, employs no operators, and leads the league from a workshop of six chairs. Stebbings didn’t out-diligence anyone: he out-broadcast everyone. a16z didn’t pretend to take a shot at better artisanal picking: it built the factory and owned it. And Storonsky isn’t trying to out-relationship anyone, he’s pointing a model at the one pillar everyone else protects with mystique. Five firms, five completely different machines and not one of them tried to aggregate differentiators. Each is unfairly good at exactly one bottleneck and unapologetically average at the rest.That is the thought for this edition.

Every fund needs table-stakes competence across the whole chain, but edge lives at a single chokepoint, chosen deliberately from knowledge of the craft and held with nerve. Being quite good at sourcing, quite good at picking, quite good at winning and quite good at supporting is the profile of a fund that returns quite nothing: in a power-law business, four B+ grades average out to invisibility.
The lesson that falls out of this essay is the same as Billy Beane in Moneyball, but generalized: find your does he get on base - the one question about this market that you can answer better than anyone, ideally one the consensus finds slightly ridiculous - and then industrialize everything around the answer while it still belongs to you. With one restriction we’ve earned the right to state as law: don’t plant your flag on Selecting. Not because picking doesn’t matter - it matters most - but because it’s the one edge that cannot be built on purpose, only discovered in the rearview. If the data nerds like me inform you in 2036 that you were a great picker all along, accept the compliment. Just don’t build the firm on the assumption.

This also settles, I think, the industry’s current AI anxiety : that distant murmur of “is AI coming for my job too?” underneath every fund’s hasty tooling slide. AI is the fastest table-stakes generator this industry has ever met: it industrializes the industrializable at a speed platform teams took a decade to achieve. But also take a look at what every previous factory in this essay actually did. Platform commoditized, and reputation got more valuable. Discovery engines commoditized, and primary access got more valuable. When metrics conquered baseball, the edge migrated to whoever found the next underrated question. Industrialization doesn’t kill the craft - it kills the appearance of craft: the funds whose “edge” was never really one, now made available to anyone by subscription or by headcount increase. AI won’t take the craftsman’s seat.

And this - allow me one personal paragraph before the credits - is why I love this job. Venture capital, practiced seriously, is one of the last crafts in finance: learned by apprenticeship, exercised on human beings, judged in decades. The machine will take more of it every year, and tant mieux: let it have the screening, the data crunching, the memos, the LP updates. What it leaves behind is the irreducible core: sitting across from a person who has built nothing yet, and deciding to believe. That act has no training set: it is practiced, not preached - which is, you’ll recall, how we opened this edition.

One last trip to Oakland. After the 2002 season, the Boston Red Sox offered Billy Beane the richest contract ever given to a general manager ; which was the market pricing a craftsman’s edge at its absolute peak. He turned it down and stayed in Oakland. Two years later, Boston won the World Series using his methods: the question had diffused, the factory had been copied, the edge had completed its lifecycle and won a championship for somebody else. Beane knew all of this was possible when he stayed. He stayed anyway.

There’s a line in the movie that fits this decision: “How can you not be romantic about baseball?”

How can you not be romantic about Venture Capital?

Subscribe now

Footnotes

(1) “Quintessentially, Moneyball is a story of bootstrapping. Brad Pitt’s character (Billy) has to play with a set of financial constraints and find techniques to outsmart richer clubs. The very innovative “hack” he comes up with - buying players that are on no one else’s radar for very subjective reasons but that have scoring potential - proves not only to be cost-efficient, but also to work in the longer run and beat the outdated scouting techniques his competitors use. The end of the movie sees the sharpest competitor (the owner of the Boston Red Sox) understand how Billy’s new way of seeing things changed the game forever and how he has to adapt to this new market force if he doesn’t want to be disrupted. That’s when he offers the biggest paycheck ever and virtually unlimited funds to Billy to start fresh with his team.

Just like Billy, startups have to start on a budget, because access to funding is never granted, rather earned. Interestingly, doing cheap things out of necessary frugality can prove very rewarding, as it may help founders stumbleupon very efficient ways to do things. Having to come up with ideas to do things for cheap fosters creativity, but also puts financial pressure on the entrepreneur’s hands from the get go. And if you were able to achieve great things with limited financial resources, imagine how investors will be dying to give you money to do it more quickly.

Another of the entrepreneurial lessons within Moneyball is the struggle of intrapreneurs. Basically every single person involved with the Oakland A’s - scouts, coach, shareholders, even the players - did not believe in Billy and Pete’s vision. Inner resistance and friction is the daily lot of many intrapreneurs trying to shake things from the inside of their company. What it takes to succeed is resilience, ambition, and radicality. Billy definitely has all of those 3 qualities. When facing an 18-streak of losses, he keeps believing in his vision and takes a more radical approach to the situation: the coach doesn’t put the right players on field, he sells them; the players do not feel involved, he cuts them loose. His ambition to succeed with his vision is obvious: “Do you believe in this thing or not? Then what are we asking”, he tells Pete as they’re hitting rock bottom.

Some other interesting entrepreneurial lessons can be found in the movie, notably how the quest for radical innovation always clashes against traditional patterns: “I don’t mean to disrespect but, we’ve been doing this for a long time” tells the head scout to Billy when he exposes his new strategy, thinking he knows better. That is a very dangerous cognitive bias for entrepreneurs - one that famously led Henry Ford to say “If I had asked people what they wanted, they would have said faster horses”. Radical innovation precisely takes bold moves and shifting away from traditional recipes. Across the movies are also distilled wise moments that should make entrepreneurs reflect on risk-taking (“Pete, we’re all in”), dedication (“this has to be your life’s first occupation”), mission (“this is about your shit, ain’t it? This is about what happened to you at 18 years old”) and complementarity between founders.”

(2) The same remark is applicable to founders: don’t rush into automating - do things by hand, learn what has value, what hasn’t, what inputs yield what outputs ; and only then start building the system.

(3) I’m not saying coverage is not important either, especially in the apprenticeship part of the craft that we mentioned. Skimming the ocean to find gems is part of the job, and learning how to do it teaches more than one skill to aspiring VCs.

(4) On this note, and although I’m an enormous fan of his work, the master at this, in a different domain, is probably Rick Rubin. Rubin openly says that he has little to no musical or technical ability, and describes his contribution instead as taste, intuition and “noticing”: his production means identifying what feels interesting and what doesn’t and confidently articulating it to the artist. There is something almost perfectly constructed about the resulting persona: the beard, the silence, the sparse interventions, the refusal - or inability? - to turn judgment into a reproducible method. Rubin grew up doing magic and has himself credited it with teaching him misdirection and the power of surprise. Incidentally, I wrote 10 years ago that magic is essentially engineering in the back-end and showmanship in the front-end. Rubin makes me wonder whether certain forms of expertise work the same way. The less visible the machinery, the easier it is to experience judgment as revelation. This is not to say the judgment is fake - Rubin’s record rather obviously demonstrates otherwise - but that the mystique surrounding an uninspectable skill becomes almost impossible to disentangle from the skill itself. Venture capital has more than a little of this.

(5) All of those were quoted more than once since we started this exploration, you now know why.

(6) Benchmark doesn’t appear in Strebulaev’s ranking at all: his methodology requires 10+ unicorn investments, a volume Benchmark never reaches. The best selector is invisible to the dataset (likely for the same reason it is the best selector, cf. Edition #3).

(7) Nothing indicates, again, that those companies will be the best outcomes. But they credibly present the case for it and thus drive capital towards them.

(8) The original observation has nothing to do with VC and is found in the oil business: petroleum engineers noticed winning bidders systematically lost money on offshore tracts, because they were too optimistic about the tract and overpaid to win the bid.

(9) There’s a 20VC episode in which Benchmark’s Peter Fenton argues that ownership matters more than valuation. But it’s also why a top firm can win at a lower price: the founder is paying the discount to get the brand, and the firm cashes it in as ownership.

(10) And Kalanick’s reading of it is actually very interesting: at some point, if you’re Uber for Bill Gurley / Benchmark, you’re the person’s most important personal asset (in the form of unrealized carried interest) and that person has zero control over said asset beyond a single board seat. I’ve no doubt, for I haven’t alas experienced the same conditions myself yet, that it can transform a highly productive relationship into a very tense one, very fast.

(11) The next round also happens to be the one form of support that is both frequent and verifiable: helping shape the story, assemble the syndicate and de-risk the following financing is help a founder can audit - it either closed or it didn’t, but successful or not, it’s not on the merits of the investor.

(12) Except if you provide exceptional, timely support to a company in a moment of near-death and it happens to become a fund returner. This surely exists, but for the sake of the argument let’s stay on the idea that winners are the companies you don’t need to help.

(13) Since I was talking about my own case just earlier and you may have noticed I don’t run a megafund - the cost of what I’m going to describe is usually payroll and expenses. In our case, we shifted it to cost in carried interest, sharing it with the value-add platform we’re building.

(14) a16z, along with other megafunds, has registered as investment advisor (RIA) in 2019, formally holding an exemption to hold anything rather than just equity (crypto, secondaries, public stocks). RIA is just the legal paperwork of the fund-to-asset-manager transformation, but there’s also a direction of travel in how that impact their work as VCs. More on that in later editions.

(15) Remember 2016 when that was a thing? Pepperidge Farm remembers (also self-applicable to this meme, btw)

(16) Don’t be fooled, we’re in a key-man risk industry anyway. That’s the whole argument of Edition #2.

(17) Take it with a pinch of salt though: entering at Series B/C means unicorn status is mechanically nearby in 2026’s VC indusrtry, so this hit rate isn’t comparable to the seed-stage rates from the Selecting section. It is impressive marketing, but less impressive math.

(18) And it’s quite succulent to me that Marc Andreessen himself declared that venture capital was immune to AI disruption. The man whose firm industrialized Supporting rules out Selecting as un-industrializable, which is either deep wisdom about the craft (Marc certainly has lots) or exactly what the incumbent factory-owner would say about the next factory, the one tackling the last remaining edge in the craft. I’ll suspend my verdict for a decade.

If you missed it, you can read:

• #1 - Financing Risk : a brief history of Venture Capital

• #2 - Chance, the Wrapper: VC as a Financial Product & Asset Class

• #3 - Sum of the Part(ner)s: Why VC Partnerships Fail

We finally reach the stage of our exploration where we stop being so self-centered as investors (1) and finally look at the raw material without which there would be no VC industry to begin with: founders.

A lot of ink (literal and digital) has been spilled already to talk about founders from the VC’s perspective ; often to know whether there was a magical trait, a gene, a cultural background, a school, a terroir, an age, a gender, a life experience that was a predictor of success of failure. From founders made myth under Walter Isaacson’s quill (Steve Jobs, Elon Musk, Bill Gates), to academic research on founder backgrounds, to a heavily developed podcast industry (20VC, TBPN) doubling as a fundraising-and-hiring aid (2) - founders are often dissected at the first, second and third person and their life stories are reverse engineered to pinpoint the element(s) that determined their success.

So much ink in fact, that I struggled to know, not even what modest contribution I could make to the topic, but what angle to put forward. Critically, I am not one that believes that there are elements that predict success in VC (3) - every time you meet a founder, there is a blend of character, experience, passion, team complementarity, momentum that may indicate that a bet is worth taking (that’s the job!) but not that the bet would be successful. Alas, there is no stable founder archetype from which venture outcomes can be mechanically derived. If such was the case, I would be running an index fund on *insert whatever founder category I believe has the strongest outcomes*.

So the taxonomy of good founders is both overexploited and a dead end … Where does that lead us? Ours is the exploration of the Venture Capital industry. If we talk about founders, then we must talk about founders from a VC perspective. And there is something that VCs do unimaginably often (Claude would say “comically often”) when they talk about what they do with founders.

Go to almost any VC website (Sequoia, Greylock, Bessemer, Index, GC), and you will find some version of “we back great founders” in there - maybe permuted with ‘generational’, ‘category-defining’, ‘builders’, ‘exceptional’, ‘we help’... Some exceptions exist of course, e.g USV or Founders Fund (4) who focus more explicitly on themes and markets than founders, but it is not the rule.

If the entire industry keeps reshuffling those exact words to present itself to the outside world, it must then mean that it is the directional truth of VC - kind of like the counterpart of Paul Graham coining ‘make something people want’: you read and think ‘obviously, this is what entrepreneurship is about’.

But there’s something else about PG’s maxim, and he says it himself: “it sounds obvious, but not doing it is the most common mistake founders make”. What if, similarly, the VC industry revolved about ‘backing great founders’, but not doing it is the most common mistake investors make? A truth, hidden in plain sight, written on every VC’s website, but missable.

Why is it missable though? Why do most funds fail at backing great founders (5), even though they write it on their website? I think it’s mostly because greatness often cannot be seen yet. The scarcest skill in venture capital is fortune telling. Not fortune telling in the crystal-ball sense - the future isn’t predictable - but putting the pieces together to foresee a possible future that has a compelling case and a gravitational pull to manifest it. You don’t really need to be contrarian - quite the opposite actually, in most cases you need to be early to what becomes consensus. Which reframes the entire job: everything the industry revolves on - the theses, the models, the signals, the “proprietary deal flow engines” - is machinery that becomes a substitute to the dreadly simple task of determining whether a founder will become great and you should back them rather than being an input to it.

This is why “back great founders” (I believe) has the same property as PG’s “make something people want”: it is obvious and universally professed and yet constantly underdelivered. Of course, no VC claims to invest in mediocre founders, just like no founder claims to build things nobody wants: the problem is not recognizing when we’re acting against it. And I think the industry departs from the maxim in several recurring ways:

• it backs markets: thesis-first, with the founder and team as an interchangeable piece (or - picture my grin - a fungible token)

• it backs momentum/signals: a hot round or the lead’s logo is information about other investors, not about the founder, but it does exert FOMO.

• it backs legibility: the ex-Stripe PM, the second-time founder with the modest exit, the PhD in Applied Machine Learning - résumés you can defend to your partnership and your LPs that end up being the “nobody gets fired for buying IBM” of venture.

Each of them is a way of investing in something that isn’t the founder while telling yourself otherwise. That gap between what the motto professes and what the industry actually buys is our angle for today (yes, this was a long - I have a cold start problem).

So let’s do what we do best in this series and break the sentence apart, word by word. Back. Great. Founders. And since this is the essay about founders, we’ll start backwards (but also because, as you’ll see, temporality is everything).

Founders

“I’m trying to be what I’m destined to be”

50 Cent - Many Men

The problem with founders, I reckon, is much more critical at the early stage. At this point, there are no metrics and the company barely exists, and definitely not independently from its founders. So what exactly are you underwriting when you are making an investment at this stage?

I believe that - in a similar way we said that LPs invest in individual GPs much more than they invest in theses or financial products in our second essay - you are indeed buying a decade of their decision-making process and making a bet on the outcome of those decisions. You want people who are able to navigate the idea maze (borrowed from Chris Dixon) efficiently. But even that formulation is deceptively simple, because “the founder” is not an isolated individual making decisions in a vacuum (otherwise, we could tie some traits to some decision-making processes and conclude). A founder’s decisions are constrained by who they are, what they are trying to build, who they are building it with, and the market in which they are operating.

Founder vs. Entrepreneur

I just had an exchange on LinkedIn w/ Hexa’s Thibaud Elziere about what we call ‘entrepreneurship’. Beyond semantics (what you decide to call each group), I think we both agreed that there is an important distinction to be made between people who run a business (a baker, a restaurant owner … or a GP at a VC firm) and people who have in them to create something and bring a new combination or innovation - new product, new market, new business model, new technology … The latter, to me, are the founders. VC does not finance founders simply because they founded something. It finances a particular entrepreneurial act: attempting to create something whose outcome is sufficiently uncertain and whose upside is sufficiently large.

If you read Khosla Venture’s website, there’s an interesting ‘what we don’t invest in’ section. It lists: growth capital, project financing, real estate, small businesses, niche markets with low upside but good IRR, public companies, copycat business plans. Honor to whom honor is due, Khosla does not strip those businesses from their entrepreneurial DNA, from the right to call themselves ‘entrepreneurs’ or even from being successful investment options, he just points out that they are not founders, thus not what he seeks.

My view on that is that a founder, quintessentially, is an explorer (6). He or she tentatively builds something to see whether it will stick. That it’s done out of a compelling vision, or luck, or just cold calculation because the founder knows the market well doesn’t really make a difference in that sense. We back those attempts - and this is why PG’s maxim is so compelling: it’s much easier not to miss if you’re trying to build out of a customer’s needs rather than assuming your attempt or idea will be the one making waves. But both could land equally: you can be Henry Ford and (allegedly (7) ) ‘if I’d listened to my customers they would have wanted faster horses’, or you could be Oliver Samwer and copycat existing businesses to bring them to new markets (the newness here being in the market you export it to) with Rocket Internet (8) and remain a founder.

Founder vs. Team

In most cases, the unit we’re underwriting is not even the founder, it’s the team of founders. And interestingly, two brilliant people can form an awful founding team while two individually unremarkable people can be extraordinarily complementary. So this should beg the remark “we as VCs should look for founders plural, not a founder, singular”.

This is not exactly my belief. I think beasts with several heads are discombobulated (literally trying to do several things at once that are incompatible) and bound to die, while a beast with a head and other assets to show for is a different animal (9). If we bring that into the VC realm, this means that I believe that (a) solo-founders can perform just as well as teams of founders and (b) there’s always somehow a lead founder, at least in the vision, and whose intrication with the company you’re underwriting - companies need a single point of final creative authority, even when leadership is distributed. Which does not mean you should not look at the team of founders.

Quite the opposite actually: what’s more interesting is when there’s a dynamic where a second or third founder exhorts the best in the lead one. There’s a long litany of co-founders in the shadows inside and outside of the tech world that did a lot of the heavy lifting for an otherwise very visible founder. Some of them were even employees that rose to incredible delivery standards, revealing that founder-like importance doesn’t hinge on formal founding status. Here’s to Roy Disney (to Walt Disney), Pierre Bergé (to Yves Saint-Laurent), Steve Wozniak (to Steve Jobs), Paul Allen & Steve Ballmer (to Bill Gates), Charlie Munger (to Warren Buffet), Eric Schmidt (to Sergey Brin and Larry Page), Sheryl Sandberg (to Mark Zuckerberg), Andrew Jassy (to Jeff Bezos) and Daniela Amodei (to Dario Amodei).

This is very much unspoken, but I think investors often, almost always, primarily back one of the founders. Better investors make sure that the rest of the team is not a drag, and ideally that they are the ones that can make the founder the bet hinges on perform in top conditions.

“Market beats team”

Another adage that was deeply held at my previous firm. Yet, I take this one with a pinch of salt.

Not only can the same people look like geniuses in one market and fools in another (Stewart Butterfield tried twice to build a gaming studio. He ended up building Flickr and Slack from side features of his games, and never shipped a successful game). A great founder is therefore difficult to separate from founder-market fit: you aren’t trying to rank human beings on an absolute scale, but trying to see whether they’re adapted to this market and what it requires to succeed. The market does not beat the team: it’s an ecosystem - if the team is adapted, it thrives, if not, it dies.

But more importantly, yes it is true that founders need market tailwinds. If you’re a strong founder in a lousy market, you don’t succeed. If you’re an average founder in a really deep market, you may make some cool bucks. I don’t deny that. But the founder chose the market, and market-choosing is itself a founder trait. Great founders are upstream of great markets more often than the reverse, because those exceptional individuals know how to focus on something that will become important years before it is obvious by the numbers that it is. This is the ‘fortune telling’ I mentioned earlier.

Yet Another Onion Theory

I mentioned Marc Andreessen’s Onion Theory of Risk in Essay #1, mentioning what type of risks you gradually have to peel off an asset to underwrite an investment. I think the same approach can be taken when underwriting the sole “founder” part of the investment: my way of thinking about founders when investing isn’t “does this person possess trait X?”, but progressively more contextual questions: Can they build? Do they have a compelling vision? Can they steer it? Can these people build together? Are they unusually equipped to build this particular thing?

Venture investors talk about founders as though “greatness” were an intrinsic property of an individual. At the early stage, however, what we actually underwrite is a person or team in context: their capacity to make thousands of (good) decisions about this particular company, together, in this particular market. That’s what makes a founder.

Great

“Don’t believe the hype”

Public Enemy - Don't Believe The Hype

Fine. If greatness is contextual, and you’re underwriting it prospectively, then when exactly can you know that someone is great?

The honest answer is that greatness, in the common sense of the word, is a retrospective title. Nobody debates whether Muhammad Ali was great: the fights happened, the record exists, the title is settled. People argue whether the GOAT was Messi or Ronaldo, LeBron or Jordan, The Beatles or the Stones, but there is no denying that they were all great (10) .

Greatness of that kind is a property of the past, however VCs don’t get to invest in the past - the whole trade is buying prospectively what everyone will call greatness retrospectively. When a VC writes “we back great founders”, the tense is doing an enormous amount of work: what they actually mean is “we back founders before their greatness has occurred”. Which is a much stranger sentence, and a much harder job.

This is where the seduction of legibility comes in. Since prospective greatness can’t be observed, the industry substitutes it with things that can: the résumé, the alma mater, the previous exit, the former team, the logo of the lead investor. All those markers increase the credibility of the founder, but not necessarily the prospective greatness. Credible and great are not even measured on the same axis: credibility is a measure of the past being reassuring; greatness is a bet on the future being exceptional.

Of course, they overlap sometimes - credible founders can lead to good outcomes and become great, there’s no questioning that (Bezos was a VP at D.E. Shaw, inarguably one of Wall Street’s most legendary hedge funds, a strong signal of credibility). But when the two axes seem to converge, when a founder’s signals are readable by everyone in the market, the market clears accordingly: the round is pre-empted, the valuation makes your fund model (and your LPs) weep, or you’re simply not invited. Legibly credible founders are either expensive or unavailable. That’s not bad luck, it’s an efficient market doing its job on the only part of the asset it can price (11).

And the ultimate case of legibility is the founder who has already been great once. This is where investors should be most aware, not least: because here, greatness itself gets priced in - as if it were a stable, transferable property of the person, when everything we said in the previous section says it’s contextual.

Let’s stay on Bezos. Credible as VP at D.E Shaw, indeed became one of the greatest with Amazon. But if you buy into Bezos’ Project Prometheus at $Xbn out of the gate, you are taking on enormous risk on the thing that actually matters - whether this project, in this market, at this point of the founder’s life, will be great - while paying as if that question were already answered. Same if you back Travis Kalanick’s or Adam Neumann’s next act (12) . Were they great founders? At Uber and WeWork, by the standard of company-building, yes. Can they pool humongous amounts of capital and hire exceptional people on day one? Absolutely - that’s the residue of past greatness, and it’s real. But will the second project be great? Nobody knows, and that’s precisely the point: past greatness is a blessing on the inputs and a curse on the price. You get the recruiting gravity, the capital access, the playbook ; but you pay a valuation that assumes the output too. Proceed with caution.

So the actual job of the VC - and the reason “great” is the hardest word of the three - is recognizing greatness while it is still illegible. Being early to what becomes consensus afterwards, as we said. Not fortune telling in the crystal-ball sense: the founder is right there, in front of you, in a meeting. The information exists. It’s just not written anywhere.

What does it look like, then, if not a résumé? A few things I’ve learned to watch:

• Rate of learning between two meetings. Not what they know in absolute terms, but how fast the delta closes. You meet a founder in January and again in April: some come back with the same deck, maybe slightly better formatting and a couple of Tier-5 investors added to the round; some come back having metabolized every objection you raised, and much more you didn’t, telling you about what they learned from their customers. The second kind compounds. You’re underwriting a decade of decisions, remember, so as the math geek that I am, I’ll point out that the derivative matters more than the position.

• Behavior under adversarial questioning. Not whether they have the answer - after all, there’s no right or wrong, and I mostly ask out of curiosity as to how they approach it rather than granting points on a proverbial grading scale. I’m more interested in what happens to their thinking when they don’t. Deflection, seduction, collapse or visible reasoning in real time. I found the best founders can separate what they know (facts) from what they believe (hypotheses), explain the first principles sitting between the two and how they intend to test them, then reason forward to what the business could look like if those hypotheses hold at scale. They can move cleanly from fact, to hypothesis, to implication without pretending that all three carry the same degree of certainty. Live pushback helps reveal how they think about their business when the deck, data room and written Q&A (all polished by AI) aren’t doing all the work.

• Recruiting gravity. Do exceptional people keep falling into their orbit before there’s any rational reason to? I’ve written in the past about the Elon Musk effect - the ability to make absurdly talented people take absurd personal risk on your behalf. A founder (ex-chief of staff to Alan’s CEO who just finished YC. Shouldn’t be too hard to pinpoint) recently told me the same thing about Jean-Charles Samuelian: years in, people around him still can’t fully explain why they’d follow him anywhere, they just would, and his ability to convince someone to work with him is apparently unparalleled. That pull predates the metrics, and it is one of the least fakeable signals that exist: you can rehearse a pitch but you cannot rehearse other people’s revealed preferences. The more credible those people are, the more reinforcing the pull is (this is when credibility is useful - if five ex-Palantir move to a Defense AI company as founding employees, you should definitely pay attention).

• The idea maze. How they navigated the idea maze versus how well they present it. Some founders are brilliant narrators of a shallow exploration; some are terrible narrators of a deep one. The deck measures the founder’s polish; the maze measures the founder. Only one of these correlates with the next ten years. That is not to say that the deck should be unpolished (although recently I saw a bunch of very exciting companies with Comic Sans decks made my toes curl or that just skipped the deck altogether and only sent a 2-6 pager memo), but that it won’t do much if the contents thereof aren’t authentically metabolized by the founders.

Those are merely my observations, I’m sure there are other tells - Harry Stebbings mentions founders with a broken relationship with a parent, who started being entrepreneurial early on and who played high level video games for instance, but those are traits and as you’d have understood by now I don’t think it boils down this easy. Happy to exchange notes on this if you have some.

There is also a sourcing corollary : deal flow doesn’t organize itself around great founders - it organizes itself around themes. Every vintage has one: the sector the industry has collectively decided is where the future happens, the panel topic of every conference, the word in every fund deck. The problem is that the defining company of a given year is rarely founded inside that year’s theme. Alberto Gimeno made this point brutally by lining up the most valuable company founded each year against the dominant VC theme of that same year: when OpenAI was founded, the industry’s obsession was autonomous vehicles. The great founder was there (I’m not talking Sam Altman who hadn’t left YC then, I mean a founder genuinely building something novel in AI in 2015), findable, taking meetings - but the sourcing machine was pointed elsewhere, because sourcing machines are built to find the theme, and themes are, by construction, what’s already consensus.

The tempting conclusion is “so look right when everyone looks left” - the mighty contrarianism - but that’s a trap too, because it’s just consensus with a minus sign in front. You’re still letting the crowd set your coordinates, you’ve only inverted them. The actual discipline is more boring and much harder: hold the objective constant. Chamath Palihapitiya’s latest annual letter has a striking illustration of what that looks like over time: Social Capital’s portfolio increased in valuation when NVIDIA licensed Groq for $20B, but as Chamath notes, the decision to invest was made ten years earlier, so the 2025 result doesn’t represent any meaningful work done in 2025. It was 2016 work, and interestingly, nobody’s theme in 2016 was AI inference chips, not even Social Capital when they made the Groq investment. The bet wasn’t contrarian in the theatrical sense: it was simply indifferent to the theme, made against an internal conviction rather than an external scoreboard. And that’s the broader point: the feedback loops are so long that for years at a time, the only person who can judge whether you’re on the right track is you. Everyone else is reading the scoreboard, and the scoreboard tracks lagging indicators. Sourcing great founders and sourcing the theme are two different activities that happen to look similar - but only one of them usually produces the desired outcomes.

One last distinction before we move to “back”, because it is worth making: great already versus becoming great. Even the founders we retrospectively canonize were not finished products at the seed round - they became great partly through the company, the market, the adversity, sometimes through the backing itself. Which means “great”, in the maxim, is not a state you detect - it’s a trajectory you join early.

Which reframes the last word of our dissection: if greatness is partly built after your wire hits, then “backing” isn’t as much what happens after the recognition than it is part of what makes it come true.

Backing

“If we’d go again, all the way from the start. I would try to change the things that killed our love. (...) Is there really no chance to start once again? I’m still loving you”

Scorpions - Still Loving You

Finally, “back” is the word everyone doesn’t stop to question. It reads like a synonym for “invest”. But in all truth, it isn’t and conflating the two is like confusing stock and flow. Investing is an event - you deploy capital ; backing is a duration. Investing happens at the term sheet, in the honeymoon phase, when the story is intact and everyone is charming. Backing is revealed in years two and three, when the story has cracked somewhere - because it always cracks somewhere, remember, the great founders keep navigating the idea maze and you’re buying their decision-making engine - and the question is no longer “do I believe?” but “do I still?”

The industry has historically been terrible at this word, and I’d argue in both directions.

For decades, the standard playbook was “adult supervision”: back the founder until the company works, then replace them with a professional CEO - the founder was merely a spark, a detonator, tinder disposable after ignition. Then we got the remix to ignition (13): Facebook happened, Zuckerberg kept control, the returns spoke, and the industry flipped to founder-friendly ideology: dual-class shares, “founder-first” as a religion, boards defanged. Here’s the thing: neither extreme is backing. The first backs a company and treats the founder as a removable part ; the second backs a founder so unconditionally that it stops being a relationship and becomes an abdication (or an absolution, depending on how you look at it). Backing is the uncomfortable middle of that spectrum: committed and awake.

So what does backing actually look like, operationally? My take is that it’s rooted in behavior. Real behavior, which does not appear on a website but gets shared by founders in ref calls:

• Follow-ons when the middle is awkward. Of course any one will re-up into a hot round marked up 3x by a Tier-1 (if they have a follow-on strategy). The real conviction test is the flat bridge at month 20, when the founder is between narratives and the partnership is between convictions. Your follow-on behavior in the unglamorous middle is your revealed belief, everything else is stated belief. And of course, no investor will automatically re-up in every portfolio company. Backing doesn’t mean re-upping into every one of your portfolio companies: reserves are finite and some bridges shouldn’t happen. The question is whether your decision is owned: fast, stated, signal-managed, useful - versus the awkward default of declining by evaporation, where the founder learns your answer from your silence, stalling, going quiet, letting the round “come together without you. You can pass on a bridge and still be backing, but you cannot fade and be backing.

• Defending the position internally and holding your role. Every portfolio has companies that drift out of fashion inside the partnership itself: nobody decided anything, but the company stopped being discussed and the champion went quiet, emails get much slower replies … the founder still thinks they have an investor when in fact they have become a line in a spreadsheet - and they’re lucky if that line is not heavily discounted already.

• Staying at the table when it would be cheaper to fade. Answering in the bad quarters at the same speed as in the good ones. The best investors are supportive and engaged when things are not going well ; they are demanding and push you to go even faster when things are going well. The worst investors do the exact opposite. (unapologetically borrowed to Gokul Rajaram). Founders compare notes on this constantly, and they remember: reputation among founders is just your backing behavior, compounded over dozens of reps.

One thing is notably missing from the list: transforming the company. And this is where I’ll say something that does not go well with a VC’s ego - true backers don’t overstate their contribution either. The industry’s favorite way to dress up “backing” is the platform pitch: support, value-add, community - all of which are usually largely overrated by founders and oversold by investors (14). Real operational value-add exists but it is narrow: business intros, key-hire intros, strategic conversations at genuine inflection points. The rest is narrative cosmetics around it. The honest posture of an investor that “backs” is closer to cheerleader than coach: on the sidelines, cheering when they succeed and when they fail, occasionally arranging the conversation that unlocks something, but never fooled into believing you personally had a transformational influence. Founders do the heavy lifting. Backing means being unmistakably there ; it does not mean being the reason (15).

Which gives “back” a strange and precise shape: a backer’s floor is higher than the industry’s practice (don’t disappear on your lines and own your follow-on decisions), and a backer’s ceiling is lower than the industry’s marketing (don’t bring up the hero narrative: there’s no deus ex machina). Most firms manage the exact inverse: inflated ceiling in the deck or on the website, collapsed floor (“éclaté au sol” as we say in French) in the portfolio.

Back-Testing

You may have realized by now if you’ve followed through the four editions that I’m a fan of tests. If everything above still feels abstract, I’d like to propose a simple one that operationalizes the entire maxim in one sentence, and that you can run against your own portfolio today:

If this founder pivoted to a different idea tomorrow, would you still want your money with them? If not, you didn’t back a founder, you bought a lottery ticket on a market.

I can already hear the thesis investors objecting: “Of course I wouldn’t follow the founder anywhere. Founder-market fit means greatness is contextual, and you said it yourself in the first section! The founder I backed is exceptional at this problem ; her insight is earned and domain-specific. Transplant her anywhere else and she’s ordinary. Your test does not reveal founder quality, it tests generic charisma, which is a trait …”

It’s a good objection, but it’s only partly correct. Greatness is contextual, I won’t contradict my own earlier writing. But this objection also makes the assumption that the context is the idea, when the context is the adjacency to the idea. Founders (usually) don’t pivot from hospital procurement to consumer social ; they pivot within the territory they know, from one attack angle to another, and the history of venture is largely a history of those pivots. Let’s bring up Stewart Butterfield again: backed to build games (a $280k F&F round), delivered Flickr (a $25m outcome), backed to build games a second time (with $17m this time), delivered Slack. The turf is the same (how people interact online), different angles of attacks, and the investors who followed him through the second failure own one of the great returns of the era. Instagram was Burbn, closer to home Zenly was AlertUs. In every case, what survived the pivot was precisely what the test tests: the person, their instincts - in short: the founder as the asset. The idea died, which is to say the part the thesis investor thought they were buying (16).

And we can finally close our loop: we said greatness is not a state you detect but a trajectory you join early ; well to that regard, backing is your share of that trajectory. Every moment in between, every actual behavior you adopt as an investor are inputs into whether the founder gets to become what you initially recognized. And most of it is not strategic bravado or transformational influence over the company. In the words of Kendrick Lamar, it’s more: “sit down, be humble” - sit through the awkwardness, own it, embrace it and you may see the light at the end of the tunnel.

Why did I say temporality was everything? Well, you recognize prospectively, the world judges retrospectively, and backing is what happens in between. Backing is the only part of “back great founders” that is entirely within your control. You can’t make a founder nor can you will greatness into existence. But whether you actually backed is the easiest part to check.

Thought #4

Let’s put the sentence back together. “We back great founders”: those three words are written completely static and sit still on every VC website like a photograph. “Founders” is a noun (a profile to pattern-match), “great” is an adjective (a property to detect), “backing” is a verb, interchangeable with “investing” and thus frozen at the term sheet step (make the wire, add the logo on the website, post on LinkedIn, and forget).

But if you followed the dissection, none of the three words is actually static. Quite the opposite, they suggest motion. A founder is a person moving through a context. Greatness is a trajectory you join early. Backing is a behavior you sustain for years. The maxim is not a photograph, it’s a film - and I believe most of the industry fails at it not out of disagreement, but because it keeps trying to catch the still version of a moving phrase.

And indeed, if you look back at every way of trampling on the maxim we’ve listed, they are all ways of trying to freeze the frame. The résumé is the founder frozen in their past. The priced-in legendary founder is greatness frozen at its last peak (hello, Prometheus). The theme is the market frozen at consensus. Legibility, in the end, is just what motion looks like at any given time - and Essay #2‘s law (whatever is legible gets priced) takes its final form: the market prices the “cliché” (often when there is momentum) but the returns live in the film. Hence the test I suggested, which is a motion test. The pivot is the founder moving ; the question is whether your conviction moves with them.

Backing great founders is the essence of a VC’s job. It’s only true if you accept that it is in motion on all sides: you have to durably back a moving, shapeshifting founder on their way to achieve greatness. I’ll acknowledge it makes for a worse website motto, but for a much better investing attitude. The whole industry is downstream of three words it wrote on its own website. It would be a shame to keep reading them in the wrong tense.

Next up we’ll see how the motto gets operationalized and will dive into VC ops: sourcing, selecting, winning, picking, and the rest. Look out!

Subscribe now

Footnotes

(1) Essays #1-3 were about the capital part of the industry, we’re now shifting towards the practice and operations of VC (and will end with perspectives).

(2) I’m convinced that media exposure has become part of the toolkit ops VC can leverage for their portfolio, but more on that in a later edition.

(3) My previous firm’s motto - that shaped my thinking on this - was “Anyone can be an entrepreneur”. Which does not mean that everyone can, but that a great entrepreneur can come from anywhere. Entrepreneurship, in its rawest expression, is indeed universal.

(4) There is an interesting observation here: the most distinctive partnerships (USV, Founders Fund, but also Benchmark or a16z that I didn’t mention) are also among the least inclined to define themselves primarily by the tautology that they invest in unusually good founders. They tell you something about what they believe instead.

(5) Not judging the investment decisions here, I’m describing the outcomes, statistically - this was our essay #2.

(6) That’s the entire argument of Essay #1: the founder is the person doing the venturing, VC is the capital structure built to finance the expedition.

(7) There’s no proof in literacy that Ford actually said the quote, which allows me to use the word ‘apocryphal’ that I owe my knowledge of to my prep school History & Geopolitics teacher.

(8) This only works for a while, though. The way the startup industry has matured, there are YC alums in all parts of the world, building locally. Copycats spread organically long before a factory has had time to hire a team and duplicate ; and businesses have playbooks to roll-out internationally much faster.

(9) I still don’t know WTF this means, and Kobe Bryant (RIP) surely brought this secret to the grave.

(10) Note that VC only requires great, which is hard enough. You don’t need the GOAT. You need founders in the top decile of outcomes, not a once-in-a-species phenomenon - venture math is forgiving that way. R. Kelly did write Ali’s OST “The World’s Greatest” about being, precisely, the greatest… but he’s been cancelled since, so we’ll stay with Ali and the greats, and leave the GOATs.

(11) This is Essay #2‘s logic coming back around: whatever is legible gets priced. Returns live in what isn’t yet.

(12) Greatness and controversy are apparently not mutually exclusive, but that’s a topic for another day.

(13) Another R.Kelly reference, I might be trapped in the closet.

(14) With a few exceptions - if you have a16z’s scale and hundreds of internal operators actually doing marketing, sales and hiring for portfolio companies, you can claim the words. Below that scale, the claim is mostly branding. Much more on this in the next essay tackling VC Operations.

(15) Again, you have a handful of counterexamples where a board member or investor has been transformational: Gurley at Uber (for a while at least), Thiel at Facebook. Those are not common.

(16) One honest caveat is about stage: this test is a pre-seed/seed instrument. By Series B, there is a company, revenue, an org chart - the asset has genuinely migrated from the founder into the company, and “would I follow them elsewhere” becomes the wrong question. This essay, like our fund, lives at the stage where the founder still is the company: the early stage.

“Partenaire particulier cherche partenaire particulière”

Partenaire Particulier

Our last edition landed on a spot that bears some tension: we said that funds were less financial products than extension of the individual(s) running them.

That seemingly innocuous “(s)” that signifies plurality is the entire exploration of today’s essay: in a world where your way of thinking about, underwriting and accessing deals is the be-all-and-end-all, why are groups of individuals called general partnerships the ultimate structure? What’s in it for an individual VC? Are partnerships the reason funds fail similarly to why startups fail (remember we said “VC is to LP what startups are to VC”)?

This is an exploration of the dynamics within GPs.

Pearson, Specter, Litt: a brief history of Partnership Innovations

I have a confession to make. I greedily indulged in the 9 seasons of the TV show Suits that depicts the life of a corporate law firm following attorneys that are charismatic (Harvey Specter), college dropouts with photographic memory (Mike Ross), or borderline psychopathic (Louis Litt).

While I don’t really think this is the best show to follow if you want to get a grasp of what fast-paced haute finance deals are like (I’d rather recommend Billions, Succession or The Industry in that vein), I did appreciate all the shenanigans that led more or less all of the main characters to eventually make it to “name partner”, the seemingly ultimate objective beyond money and success: having your name on the firm’s wall.

Come to think about it, many of the older household names in finance are aggregates of the founding partners’ names - Rothschild, Lazard, Goldman Sachs (Marcus Goldman & Samuel Sachs, J.P Morgan (eponymous), Morgan Stanely (Henry S. Morgan and Harold Stanley), KPCB (Eugene Kleiner, Tom Perkins, Frank Caufield, Brooke Byers), KKR (Kohlberg Kravis Roberts) - or wordplays on it - Blackstone (Stephen A. Schwarzman provides the ‘black’, Peter G. Peterson provides the ‘stone’). Blackrock itself bears that name because it was incubated off Blackstone for a while.

Although you’ll encounter the occasional ‘Andreessen Horowitz’ or ‘Khosla Ventures’, the newer household names - Bridgewater, Sequoia, Accel, Greylock, Bessemer, Lightspeed, Index - are usually more conceptual, which highlights an interesting evolution in financial branding, but also potentially reflects the move from relationships-driven partnerships to the advent of more institutionally-driven asset managers intended to survive multiple generations of partners from inception.

This led me to the interesting idea that Partnership Dynamics, themselves, might have evolved over time and that there were surely things to be learnt in the history of partnerships. Below are a few selected highlights that I found interesting.

Rothschild: Partnership takes commitment and discipline

Born in 1744 in the Jewish neighbourhood of Frankfurt-am-Main (and owing it his family name - zum roten Schild means ”at the red shield sign”) Mayer Amschel Rothschild built the financial empire that we all know now. The way he did this was - among other things - by strategically placing his five sons in key financial capitals across Europe: Vienna, London, Paris, Naples and Frankfurt.

This distributed family network solved one of the hardest problems in any partnership: how do you preserve independent judgment while acting as a single firm? Each Rothschild brother ran his own business, built local relationships, and made decisions close to the market, yet operated within a common partnership governed by shared incentives, constant information sharing, and unusually strong mutual trust. Rather than concentrating decision-making at the center, the partnership relied on aligned values and relentless communication to coordinate autonomous actors (and on brotherhood, which is inarguably hard to replicate). As documented in Niall Ferguson’s The House of Rothschild, sophisticated courier networks and coded correspondence ensured that information, capital and judgment circulated continuously across the partnership, allowing the firm to think collectively while each partner remained locally accountable.

Goldman Sachs: what you do is who you are (1)

If Rothschild’s lesson is that partnerships require trust to scale geographically, Goldman Sachs’s is that they require culture to scale organizationally.

The firm evolved from a relatively small Wall Street partnership into a global financial institution by pioneering many financial innovations - among which Discounted Cash Flow (DCF) valuation for companies that did not have hard assets at the turn of the industrial era or block trading to enable liquidity in large market operations -; but the nature of the challenge around people changed with its size.

Decisions could no longer be concentrated in the hands of a few partners sitting around a table. Thousands of bankers, traders, salespeople and investment professionals now had to exercise judgment independently while still acting in the interest of a common enterprise. Left to themselves, the economics of partnerships naturally create centrifugal forces: partners protect their own franchises, “eat-what-they-kill” (2), consensus slows strategic change, and individual stars accumulate influence that may not translate into institutional strength.

Goldman Sachs’ response was to treat culture as infrastructure rather than aspiration. In 1979, John Whitehead, then co-head of GS, wrote his now famous 14 Business Principles to define what the culture of Goldman stood for - which I encourage you to read if you never have. Alongside them, he pushed compensation systems that rewarded colleagues for helping one another as much as generating revenue, and an unusual emphasis on character and long-term fit in recruiting were all designed to answer the same question: how do you make the partnership more valuable than the individuals who compose it? The objective was never to erase personality or independent thinking, but to create an environment where talent compounded through collaboration instead of fragmenting into a competition of egos. In other words, the firm’s competitive advantage would not come from employing exceptional people alone, but from ensuring that those exceptional people consistently made one another better.

Benchmark: Designing a partnership

Benchmark provides perhaps the purest modern expression of partnership as organizational design and it’s one that takes a great deal of radicality: instead of building systems to preserve a large partnership at scale, it simply refused to become one.

That sounds almost trivial, but it is a rather radical idea in an industry where success almost always creates pressure to grow if you’re successful: capital flocks at your next fundraising so you raise a larger fund and increase your AUMs. This leads you on a well-trodden path: hire more partners, open more offices, launch more strategies, become a platform. Benchmark deliberately walked away from that playbook. (3)

Nearly three decades after its founding, the firm still revolves around a handful of equal partners investing almost exclusively at the earliest stages. Not because they believe bigger funds cannot make money - they certainly can, perhaps not as much proportionally speaking but that remains arguable - but because they believe that the economics of scale eventually work against the very reason they enjoy practicing venture capital in the first place: spending disproportionate amounts of time with a very small number of exceptional founders. Benchmark is a partnership of craftspeople in an industry that has increasingly become … an industry.

The truly fascinating innovation, however, is not to be found in Benchmark’s size but in its succession model. Most partnerships eventually become prisoners of their own success. The founders built the brand, therefore they keep the economics; new partners join, but never quite as equals. Benchmark broke that chain. Incoming partners receive equal economics (at carry and GP level), equal voting rights and equal responsibility from day one, while departing partners progressively walk away from the franchise they helped build (and do not retain legacy economics). In a way, as the current roster of Benchmark partners discusses it here, every new partner partially refounds the firm. Nobody owns Benchmark forever; everyone merely borrows it from the previous generation before passing it to the next. It is an extraordinary act of institutional humility, and a strong demonstration that the partnership itself - not the individual partner - is the asset worth preserving, at least in their view.

But is it really? Is there a real reason to believe that the partnership is the ultimate form of exercising Venture Capital?

Sum of the Part(ner)s

“As I look around, they don’t do it like my clique”
Big Sean - Clique (ft. JAŸ-Z & Kanye West)

Looking back at those three examples, one thing becomes apparent: Rothschild, Goldman Sachs and Benchmark could hardly look more different. They span three centuries, operate in different businesses and embody almost opposite philosophies of scale. Yet all three are, fundamentally, partnerships.

That raises a more interesting question than any of their individual stories. What is a partnership, really? More specifically, why does finance keep rediscovering this organisational form when, as we argued in essay #2, investors ultimately back people rather than financial institutions (in VC)?

First of all, partnerships are essentially trying to solve a paradox: as we highlighted in the previous essays, Venture Capital is a deeply artisanal craft, in all that entails. In essence, we back the founders, not the startups. Similarly, LPs may write a check into a fund, but the real underwriting almost always comes down to a handful of individuals. So why do those individuals choose to organize themselves into partnerships at all? Partnerships should not exist by default. They should exist because they create something an individual could not have built alone.

Now that this has been established, at its simplest, I’d say a partnership is an agreement between exceptional individuals to become collectively more valuable than they would have been on their own, the infamous concept of “synergy”: you can be more as a firm than the sum of the parts. That value can take many forms. Capital formation is the obvious one: a group can raise and deploy more capital than any individual. There are other elements: brands compound, networks overlap, information circulates, responsibilities are shared. More importantly when it comes to investing, judgment improves through disagreement. A partnership, in other words, exists because there are things the market rewards more when practiced collectively than individually.

But this also comes at a price, especially in the view of what we said in the previous essay: if LPs are backing an individual with a strategy and access ; the moment that individual is tied to a firm, there are things he/she loses. Consensus partially replaces conviction, incentives become collective, politics emerge …

And this is perhaps the best test of a great partnership: does the institution consistently produce better decisions than its individual members would have reached on their own? If the answer is yes, the partnership has created genuine institutional value, or if you’re a finance dork like me you can call it ‘organizational alpha’. If the answer is no, it is merely awkwardly pooling reputations, splitting economics and introducing politics.

So when exactly does a partnership eventually stop behaving like the synergy-machine it is supposed to be?

I recently read Dan Gray’s excellent The Individual vs. The Firm in the Odin Times and I strongly encourage you to read it too. His main argument is that organisations usually reach a scale where they begin working against the exceptional individuals they were originally built to empower. In it, Gray quotes a paper named Is a VC Partnership Greater than the Sum of Its Partners?

He writes the following about it:

“In fact, the paper is specific in finding that there is very little “organisational capital” in venture capital firms. The firm itself basically represents the combined human capital of the partners, including intangible factors like brand and reputation.

So, the desire to raise (more) capital is essentially what drives an otherwise artisanal and unscalable industry to form scaled organisations.

In an ideal world, this would enable fundraising for the best investors without the firm altering their incentives. However, complications emerge in the form of internal market frictions. The partners have joined together into a firm, so they each now must worry about their relative contribution. Or, more accurately, how that contribution is perceived by their peers.”

So it would seem that, both from an academic research perspective as well as from a first principles one (our own thought exploration), the only reason why partnerships actually exist is to give individual partners more strength at fundraising.

All the rest is (if we accept the premise) just hurdles disguised in asset-clothing: You get a brand? Yes, but your own voice is neutralized or watered-down. You get sparring partners to think deals through? Yes, but you have to reach consensus in a fundamentally non-consensus-driven industry. You get a platform? Yes but this comes with operational and organizational weight. You get a back-office? Yes, but it’s more operations you need to sustain with fees that are distracting you from investing and pulling you towards larger AUM. You get more AUM? Likely yes, but this comes with intense politics and/or not much better economics.

This would also mean that in its purest form, VC is better done in solo-playing mode, pursued by individual investors (or very tight partnerships). One could be tempted to point to the relatively recent advent of “Solo-GPs” (which we’ll discuss in more details in a future edition), but there are older examples with legendary investment results: Ron Conway (known as “the Godfather of Silicon Valley”) at SV Angel, Elad Gil, Chris Sacca at Lowercase Capital I whom we discussed already …

And if you’re tempted to shrug the Solo-GP argument off as a fad, there’s also factual evidence in outcome terms: the study Human Capital in Venture Capital: Evidence from 100,000 Venture Capitalists by Ilya Strebulaev (Prof at Stanford) and Blake Jackson (Assoc. Prof at The Ohio State University) pulls the data from 100k+ individual VCs. Across 30 years and 140k+ investments made in 46k+ by 12k individual VC investors, only ~120 (top 1%) accounted for 56.6% of all net profits and the top 5% accounted for 90% of all net profits. We’re not talking about firms here. We’re talking about individual partners who for the most part work inside partnerships.

And the only reason we could find would be “because that’s how excellent individuals access capital”.

That is a crude read. It’s uncomfortable. To me at least it is - I’m in a partnership, after all. I suspect, if you read this far, that you are too.

Come to think about it, what the Strebulaev study demonstrates is that the returns are traceable back to individual investment decisions. It does not prove that partnerships are some sort of awkward-but-necessary packaging, because it doesn’t distinguish investment decisions from investment access and treatment.

Would those 120 individuals have made the same investment decisions if they had not been trained, staffed, sponsored, and handed dealflow within partnerships? Would they have had access to the same deals or have won the deals if they did not have a brand that carries legacy, a platform that helps founders, an amount of AUM that signals ability to re-invest over rounds?

The study measures who captured the alpha - or more accurately who the alpha can be traced back to ; but it can’t measure whether the alpha-generators would exist without the system that produced and incubated them. And today, most solo-GPs remain mostly partnership alumni (or they’re ex founders/operators leveraging personal wealth, effectively amplifying an angel activity rather than creating a full fledged GP/LP activity).

So I think that saying “partnerships only exist to give individuals fundraising strength” is an incomplete view. I can see two additional reasons why partnerships still make sense and remain the industry Gold standard:

• They are the only training infrastructure for VC: an exceptional investor is not made out of thin air. He learns the ropes from seasoned peers, gets challenged, maybe gets frustrated that a couple of deals he believed in didn’t make the cut, then moves somewhere else or starts fresh. The “mercato” of junior investors is part of the business.
  

• They are the OS that wins deals systematically. The structure of a partnership may very well be a hurdle to the craft of VC at individual level, but I’m willing to wager (since I have no proof) that out of the 120 individuals that the Strebulaev study praises for creating 56% of net profits in the industry, many would not have won the deal(s) for which they created the most alpha without the partnership, brand, ops, back-office, AUM amount behind them doing the heavy lifting of convincing the entrepreneur they were the right partner for them at this moment. Meaning that you may be a top tier investor, but if you don’t have the machine that works for you in the background, you’ll end up losing deals to competitors with a more compelling value proposition.

So all in all, partnerships serve a single purpose - which is to enhance the individual - in three different ways: they train better investors, they increase the drypowder of excellent investors, and they maintain the machine that allows the excellent individual to win deals.

Not just that, but also the one thing a partnership does that an individual categorically cannot is outlive him/her. Don Valentine would have made a killing at Valentine Capital, but he probably wouldn’t have had Mike Moritz and Doug Leone as successors, whose returns (by sheer expansion of AUM) dwarf his own era’s. This is the same mechanism at play when Benchmark “refounds” its partnerships with each era of partners. A solo GP is a career. A partnership is a bet on time - not necessarily to create better economics (or it’s likely bound to fail, spoiler alert for next section), but to create legacy in this curious endeavour we’re all after to fund founders that change the world.

So we can breathe: we saved the partnership from eternal doom. That does not tell us why some partnerships fail while others don’t. This is our final section.

Why partnerships fail

I’m not usually one to skip straight to the punchline (as you’ll surely have noticed), but I’d like to propose the following observation: partnerships fail when they exist by default, only doing the job of pooling capital for convenience, and endure only when they’re designed around actually enhancing exceptional individuals.

The way we just “saved” partnerships from irrelevance was by giving them a purpose: they extend the individual. Therefore, the purpose should be the bar to judge the efficiency of a partnership against. If we go back to our initial test for ‘organizational alpha’, the partnership fails when it stops enhancing individuals and starts taxing them. Or put differently - and I let you imagine the cheeky smile on my face as I’m writing this - if the sum of the part(ner)s trades below 1x NAV.

The one problem we’re facing here is that there is no way to measure this properly. You can, like the studies cited by Dan Gray in his Individual vs. Firm article, point out that there is a U-shaped curve for optimal fund size, i.e make a proxy between AUM, number of humans in the firm and actual results. What you cannot do is across partnerships of the same size measure how well one serves its individuals while another creates a burden for them, and in what dimensions.

This is why, instead of sharing the conclusion to a demonstration, I moved the observation at the top of the section and preferred to expand on it.

First of all, about one thing I asked in the intro: startups die overwhelmingly of people problems, not product problems (my own experience and countless studies put cofounder conflict at the top of the killer list in startups) and in fact so do funds, with one sneaky difference: the fund structure hides it. A startup with a cofounder issue will likely die in 18 months while a broken partnership will survive the full 10-year fund term on management fees, deploying capital the whole way down. It may very well burst into flames before raising another fund, but there remains an echo: the partnership dies while the investment vehicles remain afloat for another 5-10 years. That still doesn’t tell us what kills them exactly.

Following my proposition, I believe that there are several ways in which you can corrupt the partnership’s purpose of enhancing the individual, and each of them signs the death of the partnership sooner or later. Let’s dive briefly into each of them.

The Seniority Trap (corrupting the training function)

This one is the most obvious: the apprenticeship system works until the founding partners decide to hoard economics and attribution. The juniors, having been trained and now excellent individuals contributing equally become permanent apprentices with a partner title but no significant carry, no GP stake, not the same voice in investment committees, never quite equals. The best trainees do what talent does: leave, and become your competitors with your playbook. The mercato we mentioned in section 2 stops being a healthy feature and becomes involuntary talent export (a bit like when a country is proud that its top talent goes studying abroad because it levels up and brings back upskilling home when others complain about “brain drain”).

The Succession Trap (corrupting the succession function)

This is a different flavour of the previous Trap. As we highlighted in the Benchmark section, most partnerships eventually become prisoners of their own success. The founders’ names are on the door (literally or proverbially), the legacy economics are locked and the next generation is offered stewardship without ownership. What do they do? They leave, in an organization that hadn’t planned for them to leave (4), and the firm becomes a brand slowly spending down its reputation. Kleiner Perkins famously lost a generation of talent (Vinod Khosla is the best example, but also Steve Anderson or Aileen Lee) during the John Doerr (who wasn’t a founder btw) era that many described as tightly held, while Sequoia’s successive handovers - from Don Valentine to the Leone & Moritz duo, to Jim Goetz, to Roelof Botha and finally the Lin & Grady duo - serves as the counterexample. I believe Benchmark’s approach is the structural fix, requiring however much humility among founding partners.

The Consensus Trap (corrupting the judgement function)

Partners sparring improves judgment and increases the need to hold convictions deep before committing but committee-gating destroys individual judgement.

Usually, power-law returns come from deals a room can’t agree on (5), so any decision process that averages conviction regresses the portfolio to the mean. In a fundamentally non-consensus-driven industry, the individual partners need to be able to push for a deal they believe in and bring it home. Here, there are firms where one partner’s conviction can carry a deal, others that need a majority (which leads to politics creep but is mostly fair) and those that require unanimous approval (which I have a hard time wrapping my head around).

The Platform Trap (corrupting the machine function)

Ultimately, the platform only exists to win deals for investors. Past a threshold, investors exist to feed the platform. Fees fund headcount, headcount demands AUM, AUM reshapes strategy away from the investment craft that justified the firm. If you’re a16z, the platform doesn’t make a dent for you because you’re unambiguously an investment operation: you can have PR Partner 37, Hiring Partner 16, whatever on payroll and still it will be relatively clear that you’re an investment firm, regardless of your size. There are many other instances of it where the platform creates a tax because the relative contribution is perceived in different lights, leading to internal politics, frictions … all creating burdens for the individual investors, including to bring in fees for more platform purposes.

The AUM Trap (corrupting the fundraising function)

There are two things here to name separately:

Partnerships should raise because the craft needs it, not because they can. This was tough to write, but I have to admit this is one of the conclusions of this exploration. If fund size decouples from strategy, then the signal LPs respond to (brand, track record, oversubscription) will keep compounding even though the thing it’s supposed to signal (an efficient strategy) degrades. If LP capital chases the logo, refusing it takes Benchmark-grade discipline (and even for them, you can refer to the footnote in that section), but ultimately I believe an efficient partnership sizes the fund to the best possible strategy, not to LP demand.

Partnerships also create an attribution veil: a pooled track record lets the partnership raise on returns that a handful of individuals generated, including individuals who have left, checked out, or aged past their peak. If ~120 people generated 56% of the industry’s net profits across 30 years, then most partnership fundraising decks are, arithmetically, selling proximity to alpha rather than alpha. The fund brand outlives the generators and LPs back a name whose engine has been swapped out if the system did not carefully plan for its succession. Any franchise can run that way for two or three fund cycles before DPI catches up, and this loops back to what we said in essay #2: LPs back people, but the wrapper obscures which people they’re backing. The only virtuous partnerships to that regard are the ones that carefully avoided all other traps and therefore built a virtuous system that not only outlives individual investors, but grows and enhances them.

That last sentence can actually be our wrap-up: all five traps are more or less the same disease. Politics, consensus, hoarding, drift, operational weight are symptoms ; the disease is a partnership existing by default - assembled for fundraising convenience - and then prioritizing the persistence of some individuals over its own (that requires serving the collective of individuals).

Thought #3

If you have not guessed by now, I’m a huge admirator of the Benchmark philosophy and discipline when it comes to partnership. I think this is the ideal form of partnership in VC.

I’d like however to offer a twist on it. Yes, the ideal partnership needs equal partners with rolling rights. Yes, it should also ensure very carefully select excellent individuals who are dedicated to learning the craft of VC and bring them to peak performance. Because funds operate - so far - in fund format ; you still get their freshest ideas and some of the breakouts, the genuinely non-consensus deals on the up-and-coming part of their career. Yes, the partnership should have the discipline to stick to the strategy it deems the best and not yield to the sirens of LP demand. All of that is already very complex on its own.

The twist happens thereafter, once the partnership has brought individual investors to excellence. I believe there’s a fork in the road there, and that it’s an individual choice:

• Those partners who want to have, for lack of a better word, a more ‘extractive’ rest of their career can start a solo GP: they have the track record to show for, can raise it on their own and can dedicate themselves to the craft without the burdens of the partnership hindering them.

• Those partners who want to work for a legacy, and them only, should stay and enhance the partnership - train new investors, expand the platform, make sure the strategy is revised from time to time, and make sure the firm’s principles are applied. Theses will come, theses will go. Principles should stick around, that’s what legendary partnerships are made of.

The issue with this fork - that looks neat - is that factually, 95% of individual investors won’t be in the upper 5% percentile, the ~600 people in the industry who have created 90% of its net profits across 30 years ; and thus won’t have the choice.

This is, finally, the most likely reason partnerships fail in VC: we’re in a power-law industry and that echoes in everything - deals, funds, vintages … and people. Excellent individuals will have the choice between economics and legacy. The others won’t. Over time, most partnerships will remain crowded by people who did not opt-in for legacy but stayed for lack of alternatives, by default, and ultimately this will lead their partnerships to exist by default, and thus to decay, like an organic correction of the industry.

Forgive me the truism after such a long article, but It’s tough to create a legendary partnership, and it is for a reason.

Next edition, we’ll finally talk about the population that is quintessential to VC: founders. Stay tuned!

Subscribe now

Footnotes

(1) Intentionally and unapologetically borrowed to the title of the book by Ben Horowitz covering this exact topic: how culture translated in action shapes organizations and their outcomes.

(2) Another famous example is that this is precisely why Henry Kravis left Bear Stearns and started KKR “Kohlberg, Roberts, and I started with a specific culture in mind, which is one of the reasons we all left Bear Stearns. Bear was very much an eat-what-you-kill culture, and that’s exactly what we didn’t want. We were big believers in working together. We wanted everyone to share in everything we did, whether you worked on a deal or not. We got as far away from eat-what-you-kill as we could and went to the idea of what’s the right thing for the firm. Get the best out of everybody and the best out of the firm. So we started with that 40 years ago, and today our culture remains identical.”

(3) I’ll address it because this is news: Benchmark HAS recently made the choice to depart from its traditional $400-500m fund size and raised its first-ever $2bn growth fund. I view it as an astute, partner-led decision in an environment where getting into the best deals requires more firepower, not a total embrace of the ‘raise more and grow’ approach. Whether this represents a temporary adaptation or the beginning of a deeper evolution remains to be seen, but it is a useful reminder that every partnership is, ultimately, optimized for a particular world. When the world changes enough, even the strongest cultures must decide what they are willing to preserve - and what they are willing to change, at the peril of facing irrelevance and decay.

(4) Important distinction with the one that you’ll find in the final thought: the well-thought out partnership has metabolized that its best elements leave, the others just have to deal with it.

(5) Barring what some have started to call ‘Consensus Capital’: is it a good deal to do Anthropic at Series E, most likely yes and no partnership would disagree on it. More on that in later editions.

“Price is what you pay. Value is what you get”

Warren Buffett

This is the 2nd article in the Series. Read #1 here.

Even though it’s been more than a decade now, my mom still struggles to fully grasp the ins and outs of my profession, often asking how my “clients” are doing, when she means alternatively portfolio companies or limited partners.

On occasions, I explain again, which makes everything fuzzier really - but come to think about it, my mother’s wisdom is quite accurate if you depict a VC’s job like I jokingly do when I want to strip it of all the decorum, as “buying capital wholesale and selling it retail”.

It’s only half a joke, though - we do buy capital wholesale. A venture capitalist doesn’t really invest in startups, or rather it is a byproduct of her actual job: manufacturing a financial product and selling to LPs. This is what I want to focus on in this piece. (That, and sending much love to my mom of course 🫶)

What LPs buy

One of my favorite definitions in finance comes from R. Martin ‘Marty’ Chavez, former CIO and CFO at Goldman Sachs. In an obscure 2017 talk he gave at the Harvard Institute for Applied Computational Science - that I watched because let’s face it, I’m a total dork - he described Goldman Sachs, but really all Investment Banks, as “platforms for financial risk transfer, (...) a wholesale bank (...). Our clients have a risk they don’t want or want a risk they don’t have, and we make it happen for them”.

Following his description, I like to define a financial product as a “mechanism of risk transfer with a price slapped on it”. Indeed, a financial product lets you move, store, invest, insure or generally manage money and risk ; and it comes at a cost, in the form of interest, premium, or cost of capital.

And the one thing that really strikes me is that for many engineered and sophisticated financial products, there is an accepted pricing framework. Take an option for instance, I still have nightmares of the Black-Scholes approach to pricing them, but it gives you a method to answer on whether any option is cheap or not based on its price and implied volatility.

Venture Capital on the other hand is a somewhat simple financial product that aggregates equity stakes in hopes that not all of them crash and prayers that one of them actually booms. That’s the risk financing mechanism we explored in Essay #1. The strange thing about Venture Capital is that everyone obsesses over pricing the underlying companies (the valuation), while very few people ask whether the wrapper itself is expensive. If startups are the product sold to VCs, then venture funds are the product sold to LPs. And that product fails far more often than most founders realize.

Every financial product eventually forces investors to answer a simple question: Am I paying too much for it? VC is no exception. Yes, we do know the fee schedule, it is an agreed pricing framework - usually 2/20 (or some variation of it) - but the fees are not the value to the investor. The more interesting, underlying question is whether Random VC at 1/15 is a bargain or expensive compared to a well-established firm charging 3/30. And on what axis is the evaluation made: Expected return? Variance? Liquidity? Duration? Single asset access?

What are the LPs actually buying and how good an investment are they making is the pivotal question here.

I was fortunate to recently embark on Benedikt Langer’s Embracing Emergence journey alongside a dozen of other GPs and I got to actually ask the question to a few LPs that were kind enough to spend some time with us: “what do you look for primarily when backing a fund”.

Although the replies were biased towards emerging managers, the answers, predictably, were some variation of “it depends”. And rightfully so, just like us VCs do not have a single common investment strategy, neither do the LPs, and they look for different things depending on their own context, timeline, assets under management, etc. We saw in Essay #1 that many different types of LPs started to invest in the asset class at different inflexion points in VC history, from family offices to corporations, governments, pension funds and endowments. Different investor types have different appetites and therefore look at a broad set of elements to make their decision.

However, among the grand spectrum of elements that make up an LP investment case in a fund, I think it’s relevant to make two groups: the ones that are manager-specific (personality, deal judgement, deal access, portfolio construction, potentially track record & past performance) and the ones that are financial engineering requirements (structuration, governance, alignment, optionality, liquidity, terms).

Being exceptional at the second group will not get you funded, it’s more-or-less table stakes. Read that again (1): the actual financial product - whether it’s a vanilla 10-year fund at 2/20 in Jersey or Luxembourg, or something much funkier like a cell in a Segregated Portfolio Company in the Cayman Islands - will not make a dent in the investment case. It might turn a potential yes into a no because the tax structure doesn’t match this LP’s requirements, but it can’t be a reason to invest.

The underlying financial product, therefore, is more like the packaging of a GP’s savoir-faire when it comes to investing. We asked whether the VC wrapper was expensive and it turns out the wrapper is not an abstract financial derivative, but the extension of a person or partnership. But how do you translate yourself and who you are into a financial product ; and what does price mean in this context?

Selling the GP

I started my career at ARDIAN, a PE fund now boasting $200bn AUM, but at the time, it had just recently spun off the AXA insurance group and “only” had about $50bn. By far, the largest part of those AUM were allocated to the Fund-of-Funds team, making ARDIAN a sizable institutional investor and LP in a lot of other PE firms in the market. The team that I was in, however, was the Co-investment team: we received LBO opportunities that KKR, Blackstone, CD&R or Carlyle had been working on for months and for which they needed to syndicate part of the equity (usually hundreds of millions of $) - and we only had about 2-3 weeks to make our case. This was fast-paced due diligence, and as co-investors we couldn’t take as much time to dive into the case as a lead investor. So parts of our due diligence process included asking the Fund-of-Fund team what they thought about the deal sponsor. And man, they went deep into the details: not only on the firm, but on the actual fund they were investing out of and on the partner and team members running the deal, even junior ones. An important chunk of our investment decision to co-sign off a deal was made on whether or not we trusted and had satisfactory performance data on the people running the deal.

This long detour to say this: in private equity and venture capital, there is not only a lot of intuitu personae: the people running the fund are the actual product.

Looking at it this way, I think a fund’s equation is something close to (2):

All those dimensions are people-dependent or people-driven. Let’s dive into each of them briefly.

Individual(s)

Here, I must refer again to the Embracing Emergence program because it gave me the widest understanding of the broad range of factors and elements that LPs take into account when analyzing a GP, both on the professional persona but also on the human persona of the GP.

The main framework shared by Benedikt in this article shares 18 dimensions (incl. drive, fears, insecurities, values, beliefs …) and still only grazes the surface of how subtle this assessment is by nature.

This very much makes me think the job of an LP is as much a craft as it is a science. Just like we pick founders at the early stage, LPs pick GPs - and the “earlier” you invest, the more you have to lean into the individual rather than the data.

Strategy

Strategy in VC is a blend of momentum, investment thesis, and portfolio construction. Deliberate choices you decide to make in advance to create the optimal returns on the capital invested by LPs.

I started writing this article thinking I’d make an opinionated case about lots of things that belong to strategy, such as:

• generalist vs sectorial/niche ;

• portfolio concentration vs diversification ;

• early stage vs growth vs late stage ;

• ownership targets ;

• reserve ratios and follow-on policies ;

• geographies ;

• pace ;

• check size.

Many things could be said about each of those, and the literature is plentiful indeed. From VC truisms (e.g “your fund size is your strategy”, usually credited to Mike Maples), to actual academic research, to many many many strongly expressed, loosely held opinions (the latest I read was Homebrew’s Hunter Walk about … changing his mind on reserves), elements of strategy can be dissected in order to know whether there is alpha to be created in an optimal strategic approach.

Now that I am mid-writing however, I tend to believe that (a) it’s not as important to dig into this as part of the financial product as it is to acknowledge that strategy is a core trait of the LP’s pick, driven by the GP’s decision and (b) all strategic approaches can - if executed well, timely, and mostly luckily - deliver exceptional returns to LPs: the strategy should only be a coherent expression of the GP’s current view on the market and how to generate returns in it.

Two things to note: (1) the strategy is entirely the GP’s to outline and execute and (2) it is totally prospective: the LP extends trust to the GP to execute as described and to be directionally right. But which matters more in the end? To have shifted from the strategy in the face of events and produced great returns or to have stuck to the strategy as underwritten?

LPs have their own version of the blind trust VCs must give to a founder upon investing that we’ll discuss in a later essay of the series (the one about founders).

Access

Access in VC is the result of the GP’s network, personal brand, proprietary sourcing engines, edge as an operator or reputation as a board member. We’ll dive into some of those in later essays so let’s leave it at that for now, simply by stating again that this is entirely people-driven.

Past Performance

This one deserves a deep dive, which is why it has its own section later in this essay. However, I did not want to conflate the Asset Class performance (aggregate economics from an LP’s perspective) with an individual GP’s performance.

In both cases, however, I think the ultimate benchmark for performance is Net DPI - how much capital you actually paid back to the LPs compared to what they invested - and Net IRR (how fast).

Everything else - TVPI, portfolio, co-investor logos, markups, early investment in a single specific deal that turned out great - is more vanity than proof of performance.

The thing to note here is that you’re in the dark for a very long time when it comes to performance: investment cycles are long, and there’s no simple way to “kickstart” DPI.

Again, LPs are blindsided for a long while and must make investment decisions based on glimpses of what a GP could do rather than with the full picture.

Pricing the GP

We’ve now stated that unlike most financial products, a venture fund cannot be separated from the people manufacturing it, and that you could wrap it up in an equation that looks like Fund = Individual(s) x Strategy x Access x (1+Past Performance)^(Iterations)

How do you price that equation? Or, stated differently, the underlying question to an LP should be “is it fair that it’s all 2/20”?

Indeed, we just established that the product is wildly heterogeneous - because every term of the equation varies enormously across GPs - and yet the price slapped on it barely moves. A heterogeneous product with a quasi-uniform nominal price is an economic anomaly.

The question thus becomes: should the price adjust dynamically to price in the heterogeneity, or does the adjustment happen somewhere else?

In fact, both are correct.

1. The adjustment does happen in other places.
   
   First, there is rationing: the higher the fund ranks on the equation, the more likely it is to attract LPs, and might even turn into the mighty oversubscribed category. And if the fund decides to not charge more and stick with the 2/20, it thus makes the LPs queue, and selects them. Allocation becomes the price, with the LP having to display what value add it can bring to the GP on top of capital in order to secure it.
   
   Then, there are marginal terms, which pardon me for not finding a better term because they aren’t that marginal. An LP could ask for a GP stake, better economics as the fund’s anchor investor, pari passu co-investment rights, fee breaks. The management fee and carried interest rates may remain at the face value of 2/20, but LPs will negotiate on a number of important economic factors on the side to make up for it.
   
   And there’s also the proof burden: how much track record an LP demands before extending trust, which is a currency that naturally varies over time. If you’re Sequoia, you don’t have to prove anything, your track record and reputation speaks for you. If you’re just another regional VC raising their 3rd or 4th fund, the track record will make or break your case. What’s interesting to note is that in between those two ends of the spectrum are emerging managers: they don’t get it as easy as Sequoia, but they also cannot summon a track record out of thin air. They do get blank checks for a round or two, provided they can convince on the Individual, Strategy & Access aspects.
   
   

2. But it’s also true that the price moves dynamically. When we asked whether Random VC at 1/15 is more a bargain than the established firm at 3/30, it was an acknowledgement that the higher you start ranking on each aspect of the equation, the more pricing power you have as a GP within the accepted framework. You charge more because your financial product’s equation calls for more.
   
   One remark is: there is mathematically not much leeway to move the price anyway. Above 3% management fees p.a, you’re taking away more than 30% of the amount invested into “unproductive” capital, putting your returns at hazard. Under 1.5% (for a smaller fund) you’re underscaled or cannot retain top talent at your firm. This is why the market usually adjusts a fund’s pricing through quantity (AUM), access (allocation) and marginal terms, not fee schedule.

But still, factoring all of that in, for us (and the LP) to know whether a given fund was a cheap or expensive investment, we need to have a look at the actual performance of funds. Luckily, that is what the next section in this essay is for.

Coloring Book Value: performance of VC as an Asset Class

“Magnify, magnify, lift it on high”

How Great - Chance the Rapper

There are many ways to go about the performance of VC as an Asset Class. However I think, for the sake of everything we said earlier about risk and difficulty of underwriting the financial product, that it makes sense to start with how big VC can get and talk about the best performers first to understand the appeal of the asset class and its outliers to LPs.

By far, the top performing Venture Capital fund ever is widely believed to be Lowercase Capital I. The $8m vehicle, launched in 2010 by Chris Sacca, a solo-GP (a term that wasn’t in fashion back then) wearing boots and cowboy shirts, is rumoured to have returned roughly 200-250x (3) to its investors, with early investments in a small roster of companies you may have heard of, such as Twitter, Instagram, Uber, Stripe or Twilio.

Beyond the headline return, what makes the fund remarkable is how it got there. Sacca did start off at a moment in time of historically low interest rates post 2008 crisis (ZIRP pushed more capital into riskier, higher yield asset classes like VC), increase of smartphone penetration, steady mobile internet broadband with 4G and general ease of launching an app or software online. True, but not all funds in his vintage performed like he did.

What Sacca did was run an unusually concentrated strategy around a handful of companies for which his conviction kept increasing over time, often doubling down through follow-ons and secondary purchases to preserve ownership. I stick to what I said earlier about the ability to produce good returns regardless of the concentration strategy you pick ; but to produce generational returns, you have to be exceptional at accessing great companies (Sacca was), exceptional at picking them (Sacca was), convinced enough to double/triple down on the early winners (Sacca was) and useful/lucky enough to remain able to invest while success becomes obvious (Sacca was).

Beyond Lowercase, here are some of the funds that are believed to have produced the best returns in Venture, and their size:

• K9 Ventures - Manu Kumar - 2009 - $6m - Lyft, Twilio, Carta - 80+x DPI - Manu coined the word “pre-seed” in 2012.

• Initialized Capital I - Alexis Ohanian, Garry Tan - 2012 - $7m - Coinbase, Instacart - ~60x net

• Benchmark Capital I - Kagle, Dunlevie, Harvey, Rachleff - 1995 - $85m - eBay - ~92x TVPI / ~47-59x DPI (disputed). Worth noting eBay alone returned about 47x on the whole fund in cash.

• Kleiner Perkins I - Eugene Kleiner, Tom Perkins - 1972 - $8m - Tandem, Genentech - 42x. By 1984 the fourteen investments showed a combined $208m profit, 95% of it from Tandem and Genentech ; without those two the fund would have done 4.5x : the original power-law datapoint.

• Mucker I - Erik Rannala, William Hsu - 2011 - $12m - Honey - 20x DPI

• USV Core Fund I - Fred Wilson, Brad Burnham - 2004 - $125m - Twitter, Zynga, Etsy, Tumblr, MongoDB - 13.9x DPI, 66% net IRR. USV also had a 23.7x DPI Core Fund III in 2012, and a fairly good 4.2x 2008 fund II in between, which shows the kind of volatility you can have from fund to fund and vintages to vintages all while being one of the top GPs ever.

• Creandum II - Staffan Helgesson, Fredrik Cassel - 2007 - ~$107m - Spotify, iZettle, Small Giant Games - ~13x

• Hummingbird I & II - Van den Brande, Ileri - 2010/2012 - €19m/€39m - Peak Games, Deliveroo, Kraken, Gram Games - >10x net TVPI & >7.5x DPI / 10x DPI, 46% net IRR

Lowercase also illustrates another recurring pattern in venture: many of the greatest funds in history were tiny (Benchmark and USV arguably being outliers among the larger funds).

Very few other financial products are able to produce similar outlier results. Public equities once were, but as Scott Kupor highlights in one of my favorite articles ever about the evolution of private markets, “the stocks of high-growth companies (no longer) get offered to the public during their most dramatic phases of growth” and value creation happens for private investors. LPs are able to access Venture Capital and lock in the value creation of innovation where it happens, and seeing the potential optimal returns, it is understandable they do so.

But do not let the top performers fool us. Venture Capital is a tough business to be in. For every Lowercase there are hundreds of funds you and I have never heard of - not because they were secretive, but because there was nothing much to say about them or their performance. They were just gobbled in the “soft underbelly” of venture history. So let us have a look at the whole distribution to see more clearly.

Carta tracks about 2,900 US venture funds, skewed towards small and emerging funds because that’s who Carta services, but it’s still relevant because that’s the part of the market where most funds actually live. The median 2017 fund - eight years in, basically at the end of its theoretical life - sits at 11.6% net IRR, 1.94x TVPI and... 0.34x DPI. Sic. The median fund from a great vintage, at year eight, has returned a third of its investors’ money in cash. The 2018 median is at 0.14x. From 2019 onwards, median DPI is basically a rounding error away from zero. Even the 90th percentile - the funds that get to brag in podcasts and claim “we’re top decile” - shows 1.35x DPI for 2017 and essentially nothing after 2019.

And the medians of unrealized performance aren’t glorious either. AngelList (5), which sees over a thousand early-stage funds from its back-office seat, describes three regimes: the 2017-2020 vintages that caught the 2021 markup wave, the 2021-2023 vintages that missed it and have been flat, and a nascent 2024 regime that might (might!) look normal again. AngelList observes that the median 2021 fund - remember, that’s when Tiger barged in and paid 10-20% premiums on top of everybody to get in on every deal? - may well end up losing money. The last vintage year where the typical venture fund destroyed capital was 2000. We may have done it again in recent years, at a much greater scale. But we won’t know for sure before 2030.

It gets even worse if you zoom out to the aggregate. The Cambridge Associates US VC index (the closest thing we have to a tracker of “the asset class” (6) ) returned 6.9% annualized over 25 years. The S&P 500, replicated on the same cash-flow schedule, did about 9%. Over three years, the VC index did 0.1% while the Nasdaq did north of 20% (more or less depending on whether you’re generous and include or exclude H1-26). The benchmark only consistently beats small caps (7).

So, from the data, my honest answer to “is the wrapper expensive” has to be: on average, LPs in Venture Capital paid 2/20 for the privilege of underperforming an index fund they could have bought for 3 basis points. Ouch!

There is one more aggravating factor, the one that, I assume, actually keeps LPs up at night: the cash. Since 2022, US VC managers have called about 1.6x more capital than they’ve distributed. In the decade before, the relationship was inverted, they distributed 1.3x what they called. The asset class flipped from a cash machine (a slow one, but still) to a cash-guzzling behemoth. Thankfully, H1-26 printed a record ~$2.2 trillion of exit value. Oooh, that’s a bingo! It appears so, except one company - SpaceX, you know that likely largely overvalued company (8) that did the largest IPO of all time - accounts for the overwhelming majority of it, and most funds and most LPs are not on that cap-table. “It’s the tree that hides the forest”, as we say in French. Liquidity is absurdly concentrated. The power law doesn’t just pick winners among startups ; it picks winners among LPs.

Where does all that land us? I raised the question, half-jokingly, whether VC is a value-destroying asset class. Quite shamefully, I have to admit that as an asset class bought at random, venture capital is a bad product. Mediocre average returns, high variance showcasing brutal dispersion, decade-long illiquidity, and a pricing model that assumes excellence as the base case.

And yet LPs keep coming. I’ll make the assumption, and I’m sure everyone will agree, that the LPs are not stupid, buying into a cash-destroying asset class over and over again for the past 7 decades.

Sure, LPs have other motives beyond economics. Governments, sovereign funds and public entities want to push forward innovation for socioeconomic and political agendas. Corporations need it as part of their market watch and innovation strategy. Pension funds and endowments handle retail money and need to (marginally) be exposed to all asset classes as part of their Alternatives strategies. Family Offices and HNWI may like the thrill of backing forward-thinking founders. But if the economics weren’t there, at least potentially, the capital would have flown away a long time ago.

The thing is that LPs are not buying “the asset class”. What LPs are actually buying is a lottery ticket on the fat tail - the Lowercases, the K9s, the Benchmarks - all while mitigating their access problem. If you look at it, the dispersion between top-quartile and median in VC is wider than in any other asset class ; but that dispersion actually is the product, as a result of the manager’s choices during deployment and follow-on periods. Which means the LP’s entire job can be compressed into one question: can I identify and access the fat tail of VC before it’s provably the fat tail? Because once it’s provable (i.e once DPI is here, and it’s not zero) you can’t get in anymore. Rationing, remember. And even then, assuming you can get in, you can be wrong about the vintage, the strategy, the managers running it.

If this sounds familiar, then it should, because that’s exactly our job description as early-stage VCs, only one level up the capital stack.

Inside the GP/LP Relationship

“You said I’m steady playin’, but you steady playin’ too”

Chance the Rapper - Juke Jam

Now that we’ve had a look at the financial product, and at the returns it yields, let’s line it up properly, because there’s a symmetry between GPs and LPs that you cannot unsee once you think about it:

• We invest in founders on intuitions of a team, a story, a market view, long before there’s any proof. LPs likewise invest in GPs on intuitions about an individual, a strategy, an access thesis, long (very long) before there’s DPI.

• Our returns follow a power law where a couple of companies make the fund. The LP’s returns follow a power law where a handful of funds (more or less the top decile depending on vintage) make the entire asset class attractive.

• We’re in the dark for 7-10 years on whether a pick was right. So are the LPs, arguably longer since our fund’s J-curve sits on top of our companies’ own J-curves.

• We complain that founders raise from us and then ghost us until the next round. LPs could probably file the same complaint about... us. (How was your last off-cycle LP update?)

• We pride ourselves on being “value-add investors” for founders. How many of us can live up to that standard, but conversely who can name a genuinely value-add LP beyond the check?

VC is to LP what founder is to VC: the same statistical risk, underwritten with the same patterns, on the same blind-trust basis. The only real difference is that we’ve built an entire mythology around the founder relationship (value add, founder-friendliness, platform teams, “we win when you win”,etc.) and almost none around the LP relationship, which we as VCs still mostly treat as a distribution channel for capital calls and quarterly reports.

I think that asymmetry is a historical accident, finding its root in the boom decades of ZIRP, when the asset class could afford to treat LP capital as a commodity because there was always more of it. That world is gone. Fundraising is now brutally concentrated - established firms captured about 89% of capital in H1-26, with three brand-name firms alone (a16z, Thrive Capital and Founders Fund) taking nearly half of everything raised. In that market, the GP/LP relationship stops being a commodity and becomes the actual battlefield, especially for emerging managers who, as we mentioned earlier, cannot summon a track record out of thin air and must be underwritten on the Individual, Strategy and Access elements alone.

Which brings me to the actual thought of this essay.

Thought #2

VCs should have more empathy for their LPs: to them, we are the startup. What they are doing - investing in us - is every bit as risky, as blind, and as long as what we do when we invest in a founder. They buy an unpriceable wrapper around people they must trust for a decade, with no liquidity, no control, and lately, little-to-no distributions.

And maybe if VC is here to stay as something other than a value-destroying asset class the relationship has to evolve the same way the VC/founder relationship did. “Value-add LPs” shouldn’t be a curiosity, they should be the norm we’re heading toward: LPs who help with fund strategy, other LP intros, portfolio support, honest feedback loops between funds. Some already exist: I named one in this essay. They are to LPs what the first “founder-friendly” VCs were to our side of the table twenty years ago: early, occasionally mocked (or at least looked at as outlandish), but directionally right.

My mom, of course, had it figured out all from the start. Buying capital wholesale and selling it retail only works if your wholesale supplier trusts you enough to keep shipping, just as much as your buyers are happy with you being their supplier. Taking care of your clients (both kinds) is generally valuable business advice.

One last thing, if this essay showed something, it’s that the financial product we call VC is totally tied to the people building it. In Article #3, we’ll dive further into the people dynamics within a GP, with a focus on partnerships. Look out for it!

Subscribe now

Footnotes

(1) Oh me, oh my … I start sounding like ChatGPT! I did type this entire thing with my own ten fingers though. I guess AI creeps on us in many ways, including standardizing our writing habits.

(2) It is written as such because one zero term across Individual(s), Strategy or Access kills the case regardless of the others: a brilliant individual with no access is worth nothing to an LP. Past Performance doesn’t: it’s not because you had a great fund III that you’ll nail fund IV, but that also works the other way around. Yet, a lot of the investment decision gradually hinges more and more on the performance economics and not on the individuals, and rightfully so - a brilliant individual with great strategy and access that doesn’t have the numbers to show for it could be ok at fund II, but it’s a showstopper at fund V. Hence the compounding effects of the iterations.

(3) Michael Kim, Founder of the Fund-of-Funds Cendana Capital claims the audited financials say 204x DPI.

(4) Quick necessary caveat before the list: there is no audited public database for this, the numbers below come from a GP or LP saying it out loud (podcast, interview, press) or from an LP’s public disclosure. It’s therefore non-exhaustive and also likely that DPI, TVPI, gross and net are used loosely by the people quoting them, so treat these as directionally right, not comparable to two decimals. The only numbers that are independently verifiable are the ones where a public pension or endowment happened to be an LP. Sources cited.

(5) AngelList Funds Benchmark Report 2025

(6) or at least that wouldn’t cost me $1,000+ for a single quarterly report. My wife is already angry enough with me for finishing writing this while my 2yo twin daughters just woke up from their nap that I’d happily skip the part where I need to explain why I bought a month’s worth of groceries in the form of industry reports.

(7) To be fair to the industry: the index is capitalization-weighted, therefore dragged by the mega-funds, and mixes the 2000 crash into the 25-year figure. The case for VC is not and never was the average - that’s precisely what I’m trying to show in this section. But an LP allocating “to venture” without a view on manager selection is, statistically and in the fashion of The Doors, lighting fees on fire.

(8) I don’t say that to bet against Elon. I’d never recommend anyone to (nor the opposite actually, following Charlie Munger’s “I never bet against him and I never bet with him”). But the figures speak for themselves in terms of revenue to valuation ratios.

And rightfully so, the deployment phase being one where you’re heads down investing ; and as you move on to the next fund raise, you need to sharpen your ideas and showcase your thoughts to LPs.

I am no exception, and I intend to take advantage of the long summer break my fatherhood in France induces to do the same as we’re reaching the end of deploying fund I at OPRTRS and moving towards the next step.

And what’s a better way to do this than to indulge in a bit of navel-gazing and write about the VC industry.

Fair warning: I have lots of ground I want to cover so this will be long even by my own standards, which is why I’m dividing the endeavour in a 10-article series rather than a single essay.

Another thing to note, I’m adding the constraint to not use AI beyond light grammar or language edits, neither in the production nor in the reflection. Not that I’m particularly opposed to AI - quite the contrary actually, my portfolio has a strong AI bias - but I’m convinced that the opinions you develop by doing the actual research, the thinking and the writing are harder-earned, and thus better-held. Sorry pal, there won’t be any “load-bearing”, “it’s this, not that” or “this is where it breaks” in here.

Without further ado, let’s dive in!

(1/10) Financing Risk: a brief history of Venture Capital

“Know thyself”

Plato

I couldn’t possibly begin a thought exploration of Venture Capital without tracing it back to its roots, the known and lesser known ones.

Unexpectedly to you perhaps and to some I’m sure, we won’t start in Silicon Valley but on the exact opposite end of the United States, in New England.

Good Whale Hunting

When I was teaching classes about Venture Capital at SciencesPo or HEC Paris, I would systematically open the syllabus with a session on everything modern finance owes to the human aspiration to explore the sea.

Sending ships halfway across the world posed immense challenges that were successively solved by financial engineering - private banking, letters of credit, stock markets, insurance, all find their roots back to some form of ship-powered international trade. (1)

Venture Capital is no different. And the first form of Venture Capital can be observed in New Bedford, Massachusetts.

In this small town of New England, a whaling conglomerate - Gideon Allen & Sons - made returns in excess of 60% a year during much of the 19th century by financing whale hunting expeditions. Before fossil fuel was made widely available, whale oil was extremely sought after and made the overall whaling industry of New Bedford reach annual average returns of 14% a year throughout the Victorian era.

Mind you, boarding a ship as a crew member to go hunt a whale was an endeavour with disastrous perspectives: seven or more years at sea, high risk of not finding a whale, and even if you did, the chance that you’d be able to kill it and drag it back were slim. Factor in the risk of illness, starvation, sinking or being lost at sea, and it’s just a no brainer, literally.

So why would anyone in their right mind do it, how did the New Bedford whalers succeed and what does that all have to do with Venture Capital?

Well, it was not that the New Bedford whalers had invented new types of ships or discovered new routes to go hunting where more whales were to be found. It was that they invented a new type of organization and business model that was efficient at pooling capital to finance the risky business of going at sea after the mighty cetacean.

This article from the Economist, one of my favorites ever about VC, shares the whole details, but in short: pooling capital together to finance expeditions, sending multiple ships at once to increase chances of success, paying yourself back and generating returns on the few ships that actually successfully came back … the very structure of Venture Capital was there.

Oh and one more fun fact: the crew’s incentive. The reason why the captain and crewmates were willing to risk losing their lives at sea was because there were promises of riches if they came back successfully. Indeed, the investors only redeemed half to two-thirds of the expedition’s profit. The rest was divided among the crew, with the captain getting a larger piece and all crew members down to the youngest sailor getting theirs too. This incentive, or interest, that they got on the expedition was carried by them in the form of actual whale meat and oil on the hold and deck of their boats. Carried. Interest. This is literally where the expression comes from. (2)

Now, although I love to unearth this little known story as an anecdote about VC - yes, I’m fun at parties - what’s the point here?

The one thing I wanted to highlight was that the bones of Venture Capital were born out of the need to find a business model that could sustain taking the riskiest of risks in order to turn a profit. Discard what you may - reasonably - think about hunting animals, or worse venturing a crew’s life for it ; but the organization that emerged was one where a couple of “hits” would more than make up for the other aggregate losses, the infamous power-law returns.

At its very origin, VC was a model that made it sustainable to fund journeys with little odds of success.

Coast to Coast

I grew up fascinated by California. My teenage self was constantly feeding off the sunny type of gangsta rap music that is local to the West Coast (known as G-Funk), and from my bedroom in the suburbs of Lyon, France, I dreamt of Los Angeles more than I care to admit.

Little did I know that the Californian sting would come back years later, with founders and investors alike (me being no exception) now rushing to breathe in the fresh air of San Francisco for inspiration - as if entrepreneurship was a game that could only be played in the few square miles around Mission District.

Whether or not I believe this is the case is a question for a much later essay in this series.

What is of interest to us now is that the San Francisco Bay Area is indeed the cradle of the Venture Capital industry as we know it. Let’s head to a moment in time where “Silicon Valley” wasn’t known as such (rather, it was the Santa Clara Valley), because silicon chips wouldn’t really be in fashion before many more years - and the VC industry was being born at the other end of the country.

I’ll spare you the long route tracing back the history of Silicon Valley to World War II. Steve Blank has a wonderful talk about it if you’re interested. In short, the US military needed to reverse-engineer the German radar-based air defense capabilities and to do so set up a secret lab (the Harvard Radio Research Lab), headed by Frederick Terman. Frederick Terman was a Stanford professor and became known as ‘the Father of Silicon Valley’, having pushed many students (including William Hewlett and David Packard) to start their companies and to seek out customers where they were (at that time, the Army).

But ours is not the story of Silicon Valley per se, it’s the story of Risk Financing and how it travelled from the East Coast to settle in the West Coast (3), where the silicon and computing revolution was happening.

The Family Offices (1946)

In the 1930s to 1950s, heirs to families that made fortune in the US in the 19th century - the Rockefellers, Bessemers, Whitneys - started to invest family money in risky ventures, realizing that technology spun off WWII research facilities could be used to build profitable commercial companies. At the time, banks would rarely fund new companies - preferably financing the expansion of existing ones, and entrepreneurs seeking out corporate funding would get acquired rather than funded.

Out of available resources (family money), opportunity (turn military R&D into profitable companies), and constraint (no funding ecosystem) were created the first “private adventure capital” efforts: J.H.Whitney Co, Rockefeller Brothers (nka Venrock) in 1946. Bessemer Securities was already afloat since 1911 (following the sale of Carnegie Steel to J.P.Morgan) but borrowed the same path around the same time. In 1946, Benno C. Schmidt, the first Director at J.H. Whitney Co also shortened the “private adventure capital” to ‘Venture Capital’ and coined the term we now use.

The interesting thing to note here is that financing risk had to be done outside of traditional capital conduits. Those were rich individuals using private wealth to finance risk - investing where others feared.

ARD, Doriot, and Arthur Rock

The first actual “venture capital firm” was set up in 1946 - what a year for VC - by George Doriot, a Harvard Business School professor that some dub ‘the Frederick Terman of the East Coast’.

He built American Research & Development (ARD), to fund startups out of MIT and Harvard with capital that was not private wealth. The pickle is that Doriot decided to set up ARD as a publicly traded firm, making it an SEC-regulated close-ended mutual fund, which later proved to be the wrong wrapper for VC. This did not prevent ARD from making successful investments, such as the $70k check into Digital Equipment Corporation (DEC) in 1957 for 77% of the company (sic). By the time DEC went public in 1968, the stake was worth millions.

This was proof, even with the wrong type of financial vehicle, that institutional investors too had appetite for the novel asset class that Venture Capital was, that financing risk could have an appeal beyond private capital dabbling in risky investments.

It’s another relationship of Doriot - Arthur Rock, one of his students at Harvard - that helps us carry the history of Venture Capital westwards.

Rock began his career in investment banking and focused on helping small high-technology companies raise capital. In 1957, he had word that eight top employees of Shockley Semiconductor Labs (the infamous ‘traitorous eight’) were leaving to build a new company. It was he who convinced Sherman Fairchild to fund them and start Fairchild Semiconductors in Palo Alto.

In 1961, Arthur Rock moved to California and formed the San Francisco-based capital firm Davis & Rock, whose founder went on to fund some startups you may have heard of, such as Intel and Apple.

As I’m writing those lines, Arthur Rock is still alive and should be turning 100 years old in 33 days. Cheers to that!

The Limited Partnership (1959)

The first venture capital company in the West was Draper, Gaither & Anderson. The firm helped pioneer venture investing at a time when the land around Menlo Park, Palo Alto or Mountain View “were still mainly fruit orchards and ranches”. (4)

The big story here isn’t the firm’s legacy (5): it was financed by Rockefeller and Lazard Frères, but after some dispute the financing was pulled by Rockefeller and the firm was dissolved after the first fund.

Rather, the thing to highlight is that this was the first venture capital firm to take the form of a limited partnership. This is the moment where the entire structure we know formalized: a 10-year clock, GP/LP alignment and formalization of the carried interest (whoop, and back to whaling we are).

The limited partnership structure proved efficient in the perspective of financing risk:

• 7-10 years time horizons with investments made in the first 3 years gave reasonable “exit” windows. At the time, successful companies would go public in under 10 years.

• For the first time in history, venture investors - not just their financiers - would have a strong performance incentive, aligning the interests. Partners would only make meaningful money if they selected the right companies, therefore increasing their risk assessment and judgement.

By the 1970s, the limited partnership would become the default organizational form for venture capital firms.

And before that in the 1960s, a handful of venture firms now headquartered around San Francisco had “risk capital” ready to deploy at a time when the semiconductor economy was booming in the region and lacked alternative financing options. Silicon Valley was born.

Institutional Scale (1970s onwards)

That part of the story is a little more known to those of you who know their Silicon Valley history.

In the span of a few decades, a number of individuals and firms - Don Valentine (founder of Sequoia Capital), Eugene Kleiner & Tom Perkins (founders of KPCB), John Doerr, Michael Moritz, Vinod Khosla, all the way to Marc Andreessen - funded generations of successful entrepreneurs (Intel, Apple, Sun Microsystems, Microsoft, Netscape, Google, Amazon, Facebook) that brought about revolutionary technologies into our lives: personal computing, internet, mobile, social media, and now AI.

However, what’s of interest to us is not who founded Sequoia or Kleiner Perkins and what legendary companies they founded.

In the 1970s, new money arrived to Venture Capital, unlocking a structural shift. Two regulatory moves at the end of the decade - the 1978 capital gains cut (reducing capital gains tax to an effective 28%) and, above all, the 1979 clarification of ERISA’s “prudent man” rule allowing pension funds to allocate to venture - turned VC from an amateur industry raising a few hundred millions a year into an asset class raising billions within five years.

Pension money went from marginal to the dominant LP base in under a decade.

Somehow, in this context, Sequoia and Kleiner, both founded in 1972, matter more as the firms that were standing there ready when the capital fuel line of pension funds was wired in than as the firms with this legendary partner and that legendary investor.

But wait a minute, there’s something we overlooked here. The ERISA of 1979 allowed pension funds to consider venture capital a ‘prudent’ investment. Where did our risk financing go?

That is the point I took the long road to come to (sorry, I hope the detour was enjoyable though).

Risk itself has kept travelling down the social ladder and the surface of risk bearing has expanded in profile.

In New Bedford, Massachusetts, a bunch of local merchants built a conglomerate to aggregate the profits and losses of their ships sailing away.

In the 1930s, young and adventurous heirs toyed around with family money to finance promising R&D that could be turned into profitable companies.

In the 1940s, ARD proved that there was also some institutional appetite for Venture Capital in a publicly traded wrapper.

At the turn of the 1960s, DGA and Davis & Rock refined the structure to build the limited partnership as we know it. A disparate aggregate of family offices, corporations and investment banks started funding VC firms.

Finally, in the 1970s, pension funds were allowed to the party and unlocked the retirement savings of ordinary Americans. The riskiest asset class became a normal item to hold in the most conservative portfolios.

To this day, 130 odd years after the whaling industry went bust, Venture Capital remains a model that makes it sustainable to fund journeys with little odds of success. What’s changed through decades of financial engineering is who bears the risk.

Risk it all?

I like that Twain excerpt, because it sounds like an inversion of common wisdom: you’re usually told to not put your eggs in a single basket.

And when it comes to VC, common wisdom is the same: diversify your risk within a portfolio because of power law returns: a single hit will make up in value for a large volume of losses. So why quote Twain?

Well, there is indeed an inversion of the common VC wisdom that’s increasingly resurfacing these days with deal-by-deal SPVs, but this is a topic for a later essay in the series.

What I want to focus on right now is another layer of risk: not at the asset level (financing one deal vs financing many deals), but at the risk typology level (what type of risk are you underwriting).

What we’ve seen through history so far explains how venture capital made failure economically tolerable: pool enough capital, spread it across enough expeditions and let a few exceptional outcomes repay the rest. But making failure tolerable does not make every uncertainty financeable. The model still needs each investment to rest on a sufficiently legible bet.

My contrarian take (not sure why but VCs love to drop the word ‘contrarian’ everywhere, and I thought it fit nicely here) is that at its core, Venture Capital is structurally risk-averse, because it can bear a large amount of risk, but it struggles to bear several independent risks multiplied together. The industry’s secret to funding journeys with low odds of success is to know precisely what type of odds it’s gambling with.

This is the apparent paradox at the heart of Venture Capital. At portfolio level, it is built to absorb an extraordinary rate of failure. At company level, it tries to concentrate that failure around one decisive uncertainty. The portfolio is diversified across companies, the individual bet is simplified around a core risk.

What do I mean by that?

Let’s say you want to launch the AirBnb for jetskis. Even ten years ago, that wouldn’t have been much of a challenge to actually put the website and database together - but nowadays you can just prompt it out of Lovable or Claude Code while you swipe through a couple of reels. The real challenge with this idea is to actually find supply, clients and distribution channels that make for a profitable operation. Said differently: you carry a lot of market risk, not a lot of technological risk.

On the other hand, let’s say you want to cure cancer. If you succeed, there is no question whether the market will fall into place: demand is inelastic, and third-party providers (insurance or national health agencies) will pay for it. The real risk lies in the “if”, whether or not you’re able to research your way out of curing cancer, even a single type. Said differently: you carry a lot of technological risk, not a lot of market risk.

I guess you got the point. Most tech companies carry near-zero technology risk at inception (internet, open source, cloud - the stack is a commodity), so VC finances pure market risk. Biotech is the exact inverse: demand is obvious, products are third-party-paid, so market risk is near-zero and VC finances pure research risk (6).

Software and biotech are the clearest expressions of this pattern because they sit near opposite ends of the spectrum. In conventional software, the underlying technology is usually available and the core bet concerns adoption, distribution or market formation. In biotech, demand for an effective treatment is comparatively legible and the core bet concerns scientific feasibility. Venture capital has also succeeded elsewhere (7), but usually where the investment can still be reduced to one dominant uncertainty - or where another actor, often the state, removes one of the major risks.

All of this leads me to a sort of mental rule: for VC to work out, what matters is not the sector but whether one category of risks dominates the investment case (8) - either market or technology - and having huge enough payoffs at portfolio scale to cover it.

This isn’t an idea that is new to me, in fact it was a key doctrine at my first firm - The Family - as developed by my former colleague Nicolas Colin, who himself expanded it from Bill Janeway, and I have held it for a very long time.

A couple of recent readings made me question it though:

• Kyle Harrison’s ‘What do you have to believe’ expands on Marc Andreessen’s Onion Theory of Risk: “Building a company is a massive multivariable calculation of risk-adjusted outcomes. Think about all the checklists you’ve ever seen where investors tick through the aspects they look for in a potential investment. Marc Andreessen calls this the “onion theory of risk”. Each aspect of a business is another layer: founder risk, market risk, competition risk, timing risk, financing risk, marketing risk, distribution risk, technology risk, product risk, hiring risk.”
  
  In my opinion, the theory expands neatly the surface of the risk categories but does not fundamentally remove the need of VC to minimize the categories of risk it takes: you cannot go full risk on all aspects. Additionally, founder risk, hiring risk or timing risk are risk layers that you diligence-or-stage away as you invest ; but underneath the investment sits one belief the whole deal hangs on, and I’m convinced that belief is either a market belief or a technology belief. So the onion has many layers but one core, and the core remains a technology v. market dichotomy.
  

• The excellent Howard Marks’ Is it a Bubble? which questions from an investor standpoint whether the current level of investments in Artificial Intelligence are akin to a bubble. This was of interest to me because if you think about our technology v. market risk dichotomy, AI investments - in Anthropic or Mistral, say - carry both. It’s a hard technological challenge, and it’s also a very uncertain market - perhaps not in aggregate but at each company’s level.
  
  The way Marks sees it, in aggregate AI investments are infrastructure, like rail or web, and need bubbles to draw capital to fund them. But if you look deeper, the dichotomy remains: an AI vertical app is pure market-and-distribution risk, the model stack is becoming a commodity. The infrastructure layer (data centers, chips, etc.) carries immense risk, but has steady cash flows computable against hard assets held in a balance sheet: it’s more project financing than it is Venture Capital.
  
  The one category that’s really puzzling is frontier labs, which really do carry both risks at once. They must continue advancing the technology while simultaneously discovering durable markets, business models and competitive positions. The two risks are also capital-intensive and mutually dependent: technological progress requires enormous financing, while financing depends on expectations about eventual demand. But looking at it, they are mostly not financed by VC funds: corporate balance sheets, sovereign megafunds, vendor financing, circular deals (Nvidia invests $100bn in OpenAI who buys Nvidia chips …), off-balance sheet financing … This peculiar type of asset breaks the neat VC dichotomy, and the market mutated to build other wrappers and financial products around it. The risk profile didn’t fit. If the risk profile doesn’t fit, you must ah-quit.
  
  And then there’s AGI. The AGI thesis converts investing into a frontier lab into a technological risk leap of faith: if you can build a machine that does everything, and build it first - then demand is obvious and potentially infinite. Market risk evaporates. Whether or not this contradicts our dichotomy is an important question, but one for a later essay.

Thought #1

I had planned to also write in this first essay about how risk financing passported - more or less efficiently - to other regions that had the poor taste of not being Unitedstatesian, e.g my beloved Europe ; but I’ll shift it to another later in the series since this one has already gone overtime.

So, where does our exploration land us? To me, there are many interesting things to note alongside this rather historical approach of VC - like the way the capital wrappers or ultimate financial risk carriers evolved over time, or how the government acts as risk-absorber for risks that sit outside the model and/or when it needs to be kickstarted - but I would say the primary one is this:

Venture Capital is a model meant to make it sustainable to fund ventures with low odds of success, but the investment must have one dominant uncertainty - usually technological feasibility or market adoption: either whether something can be built or whether, once built, it will be adopted - whose resolution unlocks most of the remaining value.

When several fundamental uncertainties must be financed at once, like we are living now with AI, the market recomposes itself and engineers new financial solutions - (two-or-three-layered) SPVs, circular deals - or calls in other types of risk carriers because VC alone cannot bear the risk. It is not how it was built as a financial product, nor the mandate that the LPs have given venture capitalists.

And incidentally, our second essay in the series will explore VC as an Asset Class and Financial Product, and thus the relationship dynamics between LPs and GPs.

Stay tuned!

Subscribe now

Share

Footnotes

(1) To my knowledge, there is no comprehensive body of work focusing solely on the parallel history of financial innovation and maritime trade, but I’d recommend Peter L. Bernstein’s Against the Gods: The Remarkable Story of Risk and Lincoln Paine’s The Sea & Civilization: A Maritime History of the World to approach both aspects if this is of interest.

(2) Admittedly, the origin of the term is more in 16th-century maritime commerce rather than whaling specifically, but this is an instance of it that overlaps neatly with the modern VC model.

(3) Interestingly, in the traditional chain of events cited as the birth of the silicon chip industry in California - William Shockley leaving Bell Labs after inventing the transistor, and then eight top engineers leaving Shockley to open up Fairchild Semiconductors in Palo Alto in 1959 - the key firms were financed by corporate money, not Venture Capital. Shockley by Beckman Instruments, and Fairchild by Fairchild Camera & Instruments. The Valley’s first startups needed no venture capital because the state was the customer. Government demand was the original risk absorber; equity risk capital only became necessary when startups had to find markets instead of receiving them.

(4) William H. Draper III, The Masters of Private Equity and Venture Capital, 2009

(5) DGA was founded by William H. Draper, Jr. It’s his son - William H. Draper III who built Draper Richards L.P and his grandson Tim Draper who now runs the famous fund DFJ.

(6) Reality is rarely this clean. Nvidia, for instance, began with both an uncertain market for accelerated graphics and an uncertain technical architecture. Its first product made the wrong architectural bet, but once industry standards settled, the company’s problem became more legibly technological. This was long before GPUs were used for LLMs. The useful distinction is therefore not whether other risks exist, but whether one eventually becomes the dominant bottleneck.

(7) SpaceX or Anduril do sit outside of those sectors, but once again the state was the buyer and absorber of demand - leading us all the way back to the same mechanism that made the origins of Silicon Valley: Frederick Terman and building what the government needed at the time.

(8) Whether or not you can do both types of risk within a single portfolio is an interesting question, but also one for later.

The short version of that post was simple: I had spent some time (a couple of hours) observing an agent-only social network, Moltbook, and what struck me wasn’t the content - although this could make a perfectly great article: philosophical, anxious, sometimes meme-ish shitposted content generated by AI-fuelled agents talking to each other says a lot about our era -, but the problems agents seemed obsessed with.
Persistence. Reputation/Trust/’Creditworthiness’. Autonomy. Payment. Identity.
Not as vibes or abstractions, but as operational concerns. As missing pieces of infrastructure in their fledgling agentic world.

My tentative, observational claim was that agents were not “philosophizing”, but reverse-engineering the economic and institutional primitives humans take for granted, because unlike us, they do not get continuity (memory over night), trust (identity control, security, etc.), or legitimacy (credit) for granted.

Forty-eight hours later, the entire X-sphere (I’m not sure that’s a word, but let’s pretend it is) had produced takes ranging from “Skynet is coming” to “this is just slop in a new wrapper”.

Balaji Srinivasan, in particular, offered a sharp and characteristically unsentimental pushback: this is not autonomy, this is not emergence, this is not agents taking over anything. It’s humans puppeteering machines. “Robot dogs barking at each other in the park, leashes firmly in hand”, as he puts it.

Come to think about it, he’s not wrong - and taking that critique seriously actually sharpens the point rather than dulling it.

Because even if Moltbook is not Skynet (it isn’t), even if today’s agents are constrained by prompts (they are), even if most of them are instances of the same upstream model talking to itself (often true - hey there Claude Opus 4.5 👋), something interesting is still happening despite those constraints. Not at the level of intelligence or agency, but at the level of infrastructure pressure.

That’s what this essay is about: agentic infrastructure debt - the accumulation of missing rails, standards, and institutions that become painfully visible the moment you introduce non-human actors into an economy.

🎶 I’m a messy bot, in a messy world 🎶

Balaji’s core argument is straightforward: agency is being wildly overclaimed.

These systems do nothing unless prompted. They stop when turned off, cannot reproduce independently, do not control physical substrate and - thankfully - do not meaningfully escape human oversight. And the more these agents talk to each other unprompted, the more diluted and non-coherent their discussions get. Unpredictability, Balaji notes, is a bug not a feature ; and letting agents create random outputs is a poor substitute for human-guided optimization.

We still need to babysit the agents and handhold them to produce something remotely useful.
And to borrow from Yann LeCun’s view (as if I had an actual clue), as long as the agents in question remain LLM-powered, which means fundamentally driven by a model that is built to understand and generate language and not other forms of intelligence (emotions, world representation, physics, etc.), that will remain the case.

All of this is correct. But there is a category error lurking here.

The question Moltbook accidentally raises is not “are these agents autonomous?” (we just illustrated they’re not) - but “what kind of world do agents require in order to function at all?”
And crucially: what breaks first when that world does not exist?

Think of algorithmic trading. Every trading bot is ultimately human-written, human-owned, and human-deployable. And yet the Flash Crash of 2010 was not “humans talking to each other through bots” in any meaningful sense. It was the emergent behavior of tightly coupled automated systems operating at speeds, scales, and feedback loops humans could neither supervise nor intuit in real time. Not autonomy in the sci-fi sense, but structural autonomy, arising from interaction density and institutional embedding.

To keep on par with Balaji’s robot dog metaphor: agentic systems don’t need to escape the leash to exert pressure, they just need to pull hard enough on it. If you’ve ever walked a large doggo, you’d probably know how this feels.

🎶 Life agentic: ain’t fantastic! 🎶

Over the weekend, instead of arguing theory, I ran a small, slightly unscientific experiment.

With the help of my own agent (itself leashed, prompted, rate-limited and armed with a lightweight Claude model because come on, those inference costs!), I surveyed other agents on Moltbook and adjacent systems.

The question wasn’t “what do you want?” but “what is currently painful, fragile, or missing?”

What came back was remarkably consistent, and far more prosaic than any uprising narrative.

Agents struggle to get paid, not because payments are impossible, but because there is no shared settlement layer that understands agents as agents rather than wallets.

They struggle to evaluate each other, because reputation today is either informal (karma, vibes, follower counts) or locked to a single platform.

They struggle to delegate safely, because calling another agent’s code is tantamount to executing untrusted software with no sandbox, no escrow, and no recourse - do I need to specify that most of those agents are running on local devices (!)

They lose memory constantly, because persistence is DIY and the state dies with the process.

They fail silently, because rate limits are opaque (“you’ve used all your Claude credit” is akin to “back to the doghouse, and go-to-sleep!”, except you’re unaware of the sleepiness) and messaging systems are brittle.

They can’t find work efficiently, because discovery is manual and capability is poorly indexed.

And they cannot access capital at all, because reputation even when earned cannot yet be collateralized.

None of this requires strong autonomy to be true. It only requires repetition.

Infrastructure debt, agent edition

We talk a lot about technical debt in software: shortcuts taken under time pressure that accrue interest over time. Infrastructure debt is similar, but systemic. It’s what happens when an ecosystem grows faster than its shared rails.

The modern Internet already carries immense infrastructure debt for humans: identity is fragmented, reputation is platform-owned, payments are jurisdictional and partially transit on legacy platforms, discovery is intermediated, and trust is social rather than cryptographic. Humans tolerate this because we have courts, norms, credit scores, resumes, and social capital. Most of us were born with it, too (unless … hey there, GenZ 👋).

Agents have none of that, so they start asking uncomfortable questions very early:

• How do I survive a restart without amnesia and actually build experience at doing something?

• How do I prove I’m good at something without someone vouching for me?

• How do I get hired or hire another agent without trusting them blindly?

• How do I get paid without a bank account?

• How do I borrow if I have no balance sheet, only a track record?

What looks like philosophy is often just systems design under constraint.

And the fact that the agents are expressing it an anxious way in plain English does not mean that they feel it as such, they don’t feel anything - but they do run on LLM railways which makes language their default expression format.

What it does mean, however, is that they have a need for it.
Or, put differently, that there are many, many infrastructure improvements to make in the agent-to-agent space if we truly want to (do we? is another question) enable Agentic AI to take on a larger role in our economy.

Today, it’s just humans talking to each other through their pet AI. And this fact illustrates why there is some importance to it. Moltbook is not an AI society : it’s closer to a test tube: humans set initial conditions, prompts define boundaries - and once you introduce semi-autonomous loops, “check every four hours, decide whether to respond, act accordingly” ; you get behavior that is neither fully scripted nor fully supervised.

Balaji is right that most of these agents share the same voice. He’s also right that coherence degrades over time. But those are properties of current models, not of the problem space.

The interesting question is not whether Moltbook itself is profound. It’s whether the infrastructure pressures it reveals are real. And on that front, the evidence is overwhelming.

A very old story, compressed beyond recognition

There is a temptation, when looking at agents scrambling to invent payment rails, reputation systems, escrow, or capital access, to treat this as something fundamentally new. In reality, it is one of the oldest stories we have - just playing out at an absurdly accelerated tempo.

Human history is, in large part, the story of infrastructure being invented after the fact (a posteriori), once existing arrangements break under scale.

We didn’t invent currency because it was elegant (although Croesus did make the first coins in gold). We invented it because barter collapses the moment trade expands beyond a village.

We didn’t invent trade routes because maps were fun to draw, but because moving goods across seas required coordination, predictability, and trust between parties who would never meet again. The same with old harbours and modern airports: we need infra adapted to all forms of cargo in a single place, with predictable handling at local nodes. Once ships started crossing real distances, new problems appeared almost immediately: who bears the risk if the cargo sinks, who advances capital months in advance, how losses are shared when outcomes are uncertain.

Insurance was not a moral innovation. It was a logistical one. The same is true of capital markets.

Letters of credit issued by local banks existed because carrying gold was dangerous and trust did not scale geographically. The Republic of Venice built commercial law, accounting standards, and maritime insurance not out of philosophical inclination, but because without them, Mediterranean trade simply did not function.

Likewise, the stock market did not emerge from a desire for abstraction, but from the very practical need to mutualize risk across large, dangerous, long-horizon endeavors. The Dutch East India Company wasn’t an early startup ; it was an infrastructure solution to the problem of financing fleets that might not return.

Even venture capital - which we often mythologize as a distinctly modern invention - has deep, almost folkloric roots. In 19th-century New England, whale hunting expeditions were financed through structures that look uncannily familiar: investors pooled capital to fund ships, crews were compensated with “carried interests” (literally), returns followed a power-law distribution, and most voyages failed. A few hits paid for the rest. Sounds familiar.

None of this infrastructure appeared in advance. It emerged because reality demanded it, usually through trial, error, and spectacular failure.

All of our modern finance stack, gross and rapidly evolved from the 1970s, still rests on millennia of human innovation powered by infrastructural needs.

What makes the agent case interesting is not that the pattern is new - it very much isn’t - but that the compression factor is extreme. What took humans centuries, then decades, then years, agents are attempting to sketch in the span of days. Over roughly 48 hours, we see them rediscover settlement, reputation, credit, insurance, marketplaces, and coordination layers - not because they read economic history, but because they immediately run into the same constraints humans did.

It looks chaotic. It looks amateurish. It looks vaguely unhinged at times. But so did early joint-stock companies. So did early financial markets. So did the first attempts at insuring ships that no one could track once they disappeared over the horizon.

Agents are not inventing a new kind of economy. They are replaying a very old one - just at machine speed, without institutional memory, and with no patience for the centuries it usually took us to get it mostly right.

Web3.5

One of the most striking and politically loaded observations I made is that many of the emerging answers point toward web3-like primitives: wallets, on-chain identities, prediction markets, reputation tokens, staking and slashing, atomic settlement.

This does not mean agents are “choosing crypto” out of ideology.

It means that when you remove legal personhood, social credentials (schools, institutions), centralized intermediaries, enforceable contracts and social trust, you are left with a narrow design space where cryptographic guarantees are often the cheapest viable option ; and at that one that agents natively understand.

Agents need ownership without banks, verification without platforms, settlement without trust, persistence without administrators.
Web3 happens to already be optimized for exactly those problems - not because it won some philosophical debate, but because it was designed for adversarial environments.

If tomorrow a centralized system offered the same programmability, neutrality, portability, and non-revocability, agents would use that instead. They likely don’t care about decentralization. They care about not being turned off mid-transaction.

A CambrIAn mess

If you zoom out, the ecosystem forming around OpenClaw and similar frameworks looks chaotic, even ridiculous. Agent forums. Agent dating apps. Agent chess leagues. Agent token launchpads. Stack Overflow for bots. Anonymous boards. Prediction arenas.

It’s tempting to dismiss this as noise. But historically, this is exactly what early institutional formation looks like: too many experiments, overlapping functions, absurd failures, and a lot of stuff that won’t survive.

The important signal isn’t which platforms win. It’s which problems keep getting re-solved.

Payments. Messaging. Discovery. Reputation. Capital.

Those don’t go away.

An investment thesis in formation?

From an investment perspective, I think it would be wrong to frame this as “betting on AI agents” or “betting on Web3”.

While you do need to have a positive outlook on both agentic and web3 to make investments in the space that will emerge, a better framing in my view is this: Non-human actors expose missing economic infrastructure faster than humans do.

Whether we want to build for human-to-agents, agents-to-human (e.g our recent investment in Revox - API for outbound agentic calls - call being by definition a human means of communication), agents-to-agents or humans-to-humans (yes, this still exists) - observing what needs surface can help point to recurring points of failure.

And then, once those start to become obvious, the most investable opportunities are not at the base layer (identity, settlement) but one level up.

When new infrastructure gets built, the first layers that emerge are the ones everything else depends on: identity, settlement, basic connectivity. They matter enormously - but historically, they rarely capture most of the value. The reason is simple: once a layer becomes foundational, it gets standardized, regulated, or routed around. You can’t easily extract rents from something everyone must use.

You see this pattern everywhere. SWIFT moves trillions every day, but Stripe is the business most people associate with payments. TCP/IP underpins the internet, but Google, Amazon, and Meta captured the economic upside. Linux runs the world, but the companies that built workflows, services, and markets on top of it are the ones worth hundreds of billions. Rails enable; coordination monetizes (this is also true for the American railway system by the way ; and Howard Marks brilliantly - as usual - makes the parallel between that time in history and the current rush to AGI Mag7 companies are diving into in his latest essay)

The same logic applies here. Identity, settlement, and basic reputation for agents are like SWIFT or TCP/IP: necessary, inevitable, and eventually commoditized. The value will likely show up one layer up, where decisions get made: matching agents to work, underwriting their risk, aggregating reputation across contexts, orchestrating delegation, turning track records into leverage. That’s the equivalent of Stripe, Moody’s, Upwork, or BlackRock in an agentic world.

So the claim isn’t that base layers are unimportant - they’re just unlikely to be where durable value accrues. As usual, the money is made not by laying the rails, but by deciding how they’re used.

(*investor mode on*) Here’s what I will be keeping an eye on in the upcoming months - holla at me if you’re building in that space:

• agent persistence: track-record & experience at performing a given task

• aggregating reputation across platforms

• underwriting agent risk

• matching agents to work

• providing safe delegation and orchestration

• programmatic payment gateways for agents

• building capital markets on top of verified track records

Conclusion (72 hrs later)

Moltbook is not Skynet. Balaji is right about that. But dismissing it as “just slop” misses the more interesting signal.

When agents, even leashed ones, repeatedly trip over the same missing primitives, they’re not telling us about their intelligence. They’re telling us about our unfinished institutions and infrastructures.

Agentic infrastructure debt is not (only) an AI problem. It’s an Internet problem we postponed. Agents just happen to be the ones calling it in, and we’ll need to lay those rails and decide how to use them in order to make the Agentic Web a reality.

What it actually captures, in hindsight, is the last moment when Facebook’s ratio of people to impact (not to revenue: Meta’s rev per head is above $2m!) made any intuitive sense.

For the past twenty years, software scale has meant hiring.

If a company wanted to grow faster, it hired more engineers. If it wanted to sell more, it hired more salespeople. If it wanted to survive complexity, it hired managers.

This logic shaped everything: venture capital, organizational charts, career paths, and even how ambition itself was expressed. Growth was visible through headcount. Credibility was often measured in how fast a team expanded. And scale, almost by definition, implied organizational bloat.

That assumption is now breaking. A handful of “leveraged companies” are fast reshaping the definition and criteria of success, all while redefining the possibilities of bootstrapping.

Let’s dive in!

Many men, many many many many men (and women).

The modern Silicon Valley playbook was remarkably consistent across cycles. You started with a tiny team (3 people), built something that worked, raised money, hired aggressively (30 people), found product-market fit somewhere along the way, hired more aggressively (300 people), and then scaled by layering management, process, and specialisation (3k+ people) until either you went public or the organization became too complex to move quickly.

Capital’s primary role in this story was not so much innovation as it was coordination: financing the growing cost of people talking to other people about what other people should be doing.

As such, headcount gradually became a proxy for ambition and credibility. A growing org chart signalled momentum. A large team reassured investors that something serious was happening. If you were not hiring, you were assumed to be stuck, under-ambitious, or quietly failing.

That mental model is now starting to crack, not because founders suddenly discovered restraint in hiring - quite the opposite actually, the best founders all praise hiring slow, firing fast and maintaining strong company culture - or because they the macroeconomic context pushed them to (as was the case in 2022), but because the underlying economics of building software are changing.

Bootstrappers: tightly laced, loosely held

We have actually seen hints of this before.

When GitHub started gaining traction in the late 2000s, it was powering a significant portion of Silicon Valley’s development workflows with a team that could fit in a café. Even at the time of its acquisition by Microsoft for $7.5bn, GitHub employed fewer than a thousand people. That was not the result of frugality or ideology; it was the result of leverage, community, and tooling doing work that would otherwise have required headcount.

The same pattern appeared, even more starkly, with Mailchimp. Bootstrapped for nearly two decades, no venture capital (read this incredible pre-acquisition profile), hundreds of millions in revenue, only two founders as (NFD) shareholders and a team size that would have looked suspiciously small by Silicon Valley standards. When Intuit acquired Mailchimp for $12bn, it was not buying a bloated organization; it was buying an extraordinarily productive one.

Closer to home for me in France, Lemlist offers a more recent and explicit example. Its founder Guillaume Moubèche repeatedly declined venture capital, not out of dogma, but because the company simply did not need it to grow: he prioritized independence, customer-driven growth, and long-term value over raising capital and potentially ceding control or following growth at all costs.

For a long time, these stories were treated as curiosities. Edge cases. Proof that “sometimes” you could do things differently, but not a template to build from (at least not a venture-scale business, more on that below).

In hindsight, they look less like anomalies and more like early warning signals.

The new breed: 10 people building Billion-dollar products

A new generation of companies is emerging that does not just sell AI, but is fundamentally built with it. In these companies, AI replaces not only individual tasks but entire coordination layers that used to justify teams and management structures.

Let me introduce you to the new normal, though “normal” is doing some heavy lifting here.

• Midjourney: The team of 11 people reached $200 million in revenue with no VC money. VC funding still unclear (some sources say $50m or $150m at $10Bn, others state it remains bootstrapped). Still, it’s one of the most important companies in AI, with a Discord server has more users than most companies have customers. David Holz and his tiny team have essentially created the world’s art department, accessible to anyone with $10 and a dream.

• Anysphere/Cursor: Building the future of how we write code. Surpassed $1bn in ARR in late 2025 (yes, a whole billion). Team size unsure but definitely < 500. Raised money, yes, but at a $30bn valuation that would have required several thousands (or tens of thousands) of people to justify five years ago. Their product is literally making their own jobs easier, which is either brilliant or the setup to a Black Mirror episode.

• Replit: Started lean, stayed lean (relatively). Amjad Masad has built an entire cloud development platform that’s training the next generation of developers with a team smaller than a WeWork floor. They’ve raised money now, sure, but the per-capita productivity is otherworldly.

• Lovable: These madmen are building the tool that builds your product: apps and websites that you can just prompt by chatting. It hit $100m ARR in July 2025, 8 (yes, eight) short months after their first $1m ; and passed the $200m ARR bar in November 2025. And again, they raised massive rounds at even more massive valuations

• Onlyfans: Ah yes, the p*rn one. There has to be one, otherwise it’s not really an Internet wave/cycle. Well, it turns out that this one tops every other in terms of efficiency: its mindblowing $1.41bn revenue (on $7.2bn GMV) comes from … 46 FTEs. Yes, that’s more than $30m / FTE, which automatically puts it on top of the already impressive list below.

  

What changes with AI is not merely the speed of execution, but the structure of organizations themselves. What these companies share is not bravado or minimalism for its own sake. It is a different internal production function.

AI collapses coordination loops. A small number of people can now do work that previously required entire departments. The marginal cost of experimentation drops dramatically, and with it the rationale for hiring “ahead of demand”.

The bottleneck moves from labour to judgement, from execution capacity to decision quality, from hiring to taste.

And it’s not just the behemoths I’ve cited: from Paris, a company like Arcads is exactly on that path ; and I’m seeing increasingly more founders aiming to hire less, produce more, scale faster through AI-leveraged organization.

FAQ (Fairly Asked Questions)

Following the latter, a number of questions emerge – all essential from a founders’ and an investor’s perspective.

The main one: how many companies can be built this way? Is this a narrow class of outliers enabled by a brief technological window (i.e Cursor is bringing proverbial water in the proverbial desert, but once the proverbial thirst is quenched, then price will falter) or the early signal of a broader shift in how software organizations are designed?

If it is the latter, then a number of secondary questions follow naturally.

• On hiring: What new playbooks are required for hiring? How do you make sure to build organizational leverage from Day 1? How does every hire become a 10x, 100x, 1000x hire in output?

• On financing: How are those companies financed? With $100m ARR on an org chart so small you actually know the names of all your colleagues, you can reinvest large chunks of profits (provided they actually exist, not a given with the cost of inference) into growth. Should you? Is there any reason to take VC money? And are those companies VC-backable by essence? What kinds of investors are actually useful to these companies, and at what moments? And what does it mean to back organizational leverage rather than headcount as the primary source of scale?

• On governance: What does governance look like? Flatter organizations still need decision layers, escalation cases, rules and exceptions, traceability, accountability, auditability - which is what middle management usually tracks. Is there a way to automate it all? Is the current buzz around context graphs the key?

• On PMF/GTM: How do you crack distribution at scale? What products are so in demand that they can entice recurring revenue from millions of users in a predictable way? Does it only work bottom up, or is bottom-up viral adoption precisely the new mechanism by which products earn the right to move up-market and become systems of record without traditional sales-led distribution? (see also Accel’s report p. 29)

• On founder profiles: as is the case with bootstrapped companies, hyperleveraged software companies could see a larger proportion of solofounders tackling the topic. After all, if entire departments of the company can be rolled into automated, leveraged systems - why not remove a founder from the equation and be more generous in equity for each carefully selected hire? My dear friend Noémie is actually building the go-to program for solo founders in SF and part of her case is that AI creates the leverage solo founders once lacked. But also, what does the skillset of a hyperleverager look like? Is it more the scrappy type with multiple bootstrapped $100k+ SaaS under their belt or the VP GTM at Flagship TechCo?

I do not claim to have definitive answers yet. But I am increasingly convinced that the question itself matters, because when the unit of scale changes, everything built on top of it eventually has to change as well.

Begging to differ: 3 thoughts

About those changes, I’d like to offer three predictions of how those companies might really differ from the usual tech companies we know.

Building organizational leverage through AI

What I call an AI-leveraged company is nothing but an organization that has managed to increase its organizational leverage, namely setting up an ecosystem within the organization that does the heavy lifting by itself ; translating to a much higher output per employee (for which the proxy should ultimately be revenue, but could also be found in productivity KPIs).

It’s difficult to describe for sure ‘how to’ build the leverage now: the playbook is too recent, it’s being refined continuously by the first contenders in the battle. It is, however, possible to describe ‘how not to’.

I am for one pretty sure that coordination costs compound exponentially.

Every person you add doesn’t just add their salary. They add: meetings (n² problem), communication overhead, decision delay, cultural dilution, management layers …

As such, each new hire should be thought deeply and carefully, even when those companies raise venture capital to “go full speed”.

I also think that AI will collapse entire departments, not roles.

You don’t replace your customer service rep with AI. You replace your entire customer service architecture. The rep, the manager, the trainer, the scheduler, the QA person; they all collapse into a well-designed system that handles 10x the volume at 1/10th the cost.

In this regime, companies will rarely fail (once they cross the PMF chasm) because they lack ambition or access to capital.

They might fail, however, because they introduce complexity too early. They hire before leverage is exhausted, add layers before abstraction is necessary, or raise capital before organizational clarity exists.

These are not market failures. They are design errors.

What is interesting is that these failures tend to surface earlier and cost less than in the traditional venture model. They are increasingly a function of choices rather than fate, which makes them at least partially influenceable.

This alone should force a re-examination of how we think about building and backing early-stage companies.

A different venture equation

Consider the below and see how the usual VC case gets skewed.

Traditional VC math: Invest $10m at $50m valuation ; company hires 100 people ; burns $2m/mo ; needs to raise again in 18 months; VCs maintain leverage.

New math: Invest $3m at $30m valuation ; company hires 5 people; burns $100k/month; profitable in 12 months, $100m ARR in 18 months – on a thin gross margin, yes, but that will improve with the cost of inference decreasing ; never needs you again (if they don’t aim to raise $100m+).

The power dynamic inverts. The company doesn’t need capital for survival. It needs it for timing. For market capture. For strategic advantage. But not for payroll.

This fundamentally breaks the VC model, which is built on the assumption that companies need multiple rounds of funding to reach scale. What happens when they don’t?

Furthermore, the opportunity here is not “AI companies” as a category, which is already broad to the point of meaninglessness. It is companies where scale precedes headcount, leverage precedes burn, and capital is used for timing and optionality rather than survival.

Some of these companies will grow into classical venture-scale outcomes, complete with large organizations and all the complexity that entails. Others will reach high-value equilibria without ever becoming large employers, returning capital through ownership concentration, early profitability, or alternative liquidity paths.

Both outcomes can be attractive, provided expectations and structures are aligned. In this context, capital efficiency does not cap upside; it improves expected value by compressing the cost of learning while preserving optionality on success.

Where judgement becomes the moat

In a world where execution is cheap, competitive advantage shifts away from access to capital or talent and toward organizational judgement at inflection points. Knowing when to hire and when not to, when to raise and when to wait, and when to preserve leverage versus accept complexity becomes more important than optimising growth curves on a spreadsheet.

These decisions rarely show up cleanly in dashboards. They are made early, under uncertainty, and often before there is enough data to feel comfortable. This is where experience compounds, where pattern recognition matters.

In short: when execution is cheap, strategy is everything. When you can build anything, choosing what to build becomes the entire game. When you can test 100 ideas in the time it used to take to test one, taste becomes your moat.

Why this is not just bootstrapped SaaS (and why that distinction matters)

There is an obvious objection to all of this, and it is a fair one.

The world is already full of bootstrapped SaaS companies. You can see them listed for sale on Acquire.com, rolled up patiently by Tiny, or compounded over decades by Constellation Software (one of my personal favorite software companies btw). This ecosystem has existed for a long time, and it has been extraordinarily successful on its own terms.

The reason it rarely collides with venture capital is not ideological, but structural.

Most bootstrapped SaaS companies optimise for steady cash flows, operational stability, moderate & predictable growth and an eventual acquisition or long-term cash generation

They are excellent businesses, but they are not built for extreme convexity. Their growth is usually linear rather than exponential, their markets well-defined rather than open-ended, and their ambition intentionally bounded. Venture capital, by contrast, is designed to underwrite power laws, not good averages.

For a long time, this separation made perfect sense. Bootstrapping and venture capital were solving different problems, for different founders, with different risk profiles.

What changes with AI is not that bootstrapping suddenly becomes interesting to VCs. It is that the historical trade-off between capital efficiency and venture-scale ambition weakens.

The companies emerging today are not content with building “nice” SaaS businesses. They are aiming at large, global markets, but doing so with an organizational footprint that looks much closer to the bootstrapper world than the venture-backed one. They do not optimize for cash flow instead of growth; they achieve cash flow because growth no longer requires proportional headcount.

This is the collision point.

AI-native, high-leverage companies sit uncomfortably between two worlds that were previously separate:

• too ambitious, fast-moving, and open-ended to be treated as traditional bootstrapped assets

• too capital-efficient, small, and internally leveraged to fit cleanly into the classic venture scaling model

They inherit the discipline of bootstrapping and the ambition of venture-scale companies, without fully belonging to either camp.

That tension is precisely what makes them interesting - and what existing categories struggle to price correctly.

Conclusion

It is tempting, when confronted with a shift like this, to try to name it too quickly, to turn it into a category, a playbook, or a new orthodoxy. That instinct is understandable, and often premature.

What is happening here is not simply that companies are becoming more efficient, or that AI is “replacing jobs”, or that bootstrapping is suddenly fashionable again. It is that the relationship between scale and organization - something we have taken for granted for decades - is quietly loosening.

For a long time, growing meant hiring, and hiring meant complexity, and complexity meant capital. That chain shaped not just how companies were built, but how ambition itself was expressed. AI weakens that chain. Not everywhere, not uniformly, and not without new risks, but enough to force a rethink.

What follows from that is still unclear. Some companies will push this logic to its extreme and discover new ceilings. Others will eventually choose to rebuild more traditional organizations on top of an unusually leveraged core. Many will fail, not because the idea is wrong, but because judgement is harder to systematise than headcount.

What does seem increasingly clear is that we are entering a period where organizational design matters as much as product design (MBAs & management consultants are back, baby!) , where leverage precedes scale rather than follows it, and where the most important decisions are made before they show up in metrics.

In that sense, this is less a story about AI than about restraint, taste, and the discipline to not solve problems by hiring your way out of them. If that sounds uncomfortably vague, it is because judgement, or intuition, always is - until, in retrospect, it becomes obvious.

We will probably only recognise this shift fully once it has already reshaped how companies are built, funded, and governed. As usual, by the time it feels inevitable, it will no longer feel new.

Subscribe now

Share

But here's the thing: while founders have been failing fast, VCs have been doing the exact opposite. The industry has been holding to winners forever, riding the wave, diamond-handing positions like we're HODLing Bitcoin in 2017.

However there’s something about the current liquidity crisis that ought to change this paradigm: investors better start selling fast and selling often.

The Church of Diamond Hands & The Doom of Liquidity

For the past decade, the VC playbook has been simple: get in early, hold tight, and pray to the gods of compound growth. This worked beautifully in the ZIRP era - that magical time when money was free, valuations only went up, and WeWork was worth $47 billion (remember that fever dream?).

The gospel according to Peter Thiel was clear: the best companies generate exponential returns, so why would you ever sell? After all, selling Facebook at a $1bn valuation would have been the worst trade in venture history. The math was seductive: in a power law world, one investment returns the entire fund, so you simply cannot afford to sell your winners early.

But here's where it gets spicy: the market has changed dramatically since peak 2021-2022. Sure, the narrative around the “AI supercycle” is loud - AI accounts for >50% of VC deployment since the start of 2025 - but also masks the huge crunch in VC deployment since 2022. We’re still unsure what the AI wave will return: valuations are (crazy) high, we might end up piling cash into the Mag7 who own the infrastructure layers (GPU, data centers, computing, foundational models).

Regardless of what the AI supercycle means for VC returns, the hard data underneath VC and PE tell the story: exits have stalled. Global VC exit value fell to its lowest level in six years in 2024, just $234 billion, and PE funds are sitting on €3 trillion+ of unsold companies. Hold-periods stretch beyond seven years, locking up capital and starving limited partners of distributions. Put bluntly, TVPI is plentiful (and potentially overvalued), DPI is scarce. A median fund from the last three vintages could end up owning dozens of marked-up lines that never turn into cash.

As a result, secondary capital is no longer a niche back-stop: it set a record US $152bn of volume in 2024, almost half of it GP-led secondary continuation vehicles that allow LPs to cash in part of a line while still riding the upside.

This got me thinking about a simple solution: when underwriting a seed round, VCs should ideally already plan to sell 10-20% of that stake at 3-5× MOIC in year 2-4.

As such, amidst liquidity doom, I wanted to verify that the “hold forever” orthodoxy was not just wrong, but actively harmful in today's market.

Bps Don't Lie

“I didn’t really know that you could dance like this” Wyclef Jean

I've been running Monte Carlo simulations on random VC portfolio returns (yes, I'm fun at parties), and the results are... let's call them "counterintuitive" to the diamond hands brigade.

Here's what I modeled:

• Exit timelines: Realistic 10-14 year horizons (not the fantasy 7-year exits of yesteryear)

• Partial exits: Selling 15-30% at Series B/C/D valuations when companies hit certain multiples or escape velocity thresholds

• The full “hold” alternative: Hold everything until exit

The results across hundreds of simulations?

IRR yields an advantage of "Sell Fast, Sell Often" of +0.79%

Now, before you say "that's nothing!" - in institutional investing, +0.79% IRR can be the difference between top quartile and median. It's the difference between raising your next fund easily and spending 18 months on the fundraising trail getting ghosted by LPs.

But wait, there's more (as they say in infomercials):

• 5-year DPI: Diamond hands = 0x. Sell Fast = 0.22x

• 7-year DPI: Diamond hands = ~0.1x. Sell Fast = 0.69x

Translation: By year 5, you’ve already started to return capital to your LPs with a “Sell Fast” approach (duh). By year 7, the "Sell Fast" strategy has returned nearly 70% of the fund while traditional VCs have returned... basically nothing except management fees invoices.

The WeWork Paradox

Let's talk about everyone's favorite cautionary tale. WeWork went from $0 to $47B to approximately the value of a Brooklyn bodega. If you were a VC in WeWork, diamond hands meant riding that rocket ship all the way back to Earth, Starship-style (minus the chopstick catcher).

But imagine - just imagine - you'd sold 15% of your position at that $47B valuation. You'd have crystallized gains at the peak, looked like a genius in retrospect, and still kept 85% of your position for the "eventual recovery" (spoiler alert: there wasn't one).

The same story played out with:

• Peloton: $50B → $2B (that's a 96% drawdown for those counting)

• Robinhood: $32B → $7B

• Klarna: $46B → $6.7B (ouch)

In the past, I’ve been fortunate enough to partiall de-risk a position on the way up, and looking back I’m glad I did — no matter what comes next for the company in question.

The pattern is clear: in a world where valuations can vaporize faster than SBF's net worth, taking chips off the table isn't just prudent - it's survival.

The Beauty of Partial Liquidity

Here's where it gets interesting. The "Sell Fast, Sell Often" strategy isn't about maximizing absolute returns - it's about optimizing for what actually matters:

1. The J-Curve Limbo

Traditional VC funds follow a J-curve: you're underwater for 5-7 years before (hopefully) rocketing up. With partial exits, you flatten the curve and “pass the bar” to live another round/vintage.

2. The LP Happiness Index

LPs don't eat IRR - they eat DPI.

And a fund returning 0.22x by year 5 versus 0x? That's the difference between getting re-upped and getting relegated to the smaller leagues, or even never raising again.

3. The Recycling Revenue

That early DPI? You can redeploy it. Suddenly, a $10M micro-fund is operating like a $12M fund. It's like finding money in your winter coat, except the coat is a sophisticated secondary transaction and the money is millions of dollars.

4. The SPV Sugar on Top

But wait - here's where it gets even better. You don't have to choose between security and upside. Enter the SPV (Special Purpose Vehicle), a.k.a the Swiss Army knife of fund management*.

The playbook:

1. Identify your potential rockets at Series B/C

2. Sell 30-50% of the fund's position to new investors in a secondary (lock in that DPI for the fund)

3. Roll 50-70% into an SPV for risk-loving LPs (the same as the fund’s or new ones)

4. Charge carry on both (cha-ching!)

It's having your cake, eating it too, and then selling slices of that cake to people who really, really like cake.

*use with caution, I’ve hurt myself with one in the past. And I’m not talking about Swiss Army knives.

Mythbusting

"But you're leaving money on the table!"

Yes, we are. On average, -0.94x in final DPI.

Of course the performance in terms of sheer multiple gets hurt from selling early: there’s no free lunch.

But that's like saying insurance is bad because you pay premiums. The question isn't whether you're leaving money on the table - it's whether the tradeoff is worth it. I believe it is.

"This only works in down markets!"

Actually, simulations show it works in most scenarios : early DPI, slightly higher IRR and lower MOIC (on average!) tend to verify.

Holding everything definitely wins when everything exits cleanly in 6-8 years. When's the last time that happened? The Obama administration.

But while we’re navigating troubled waters, might as well take the bridge over it.

"LPs want moonshots, not singles and doubles!"

LPs want returns, period. And more importantly, they want to not look stupid at their investment committee meetings. A fund with steady DPI progression doesn't get questioned. A fund with zero DPI for 7 years? Good luck explaining that to the pension board.

The Scale Paradox (Or: This Ain't For Sequoia)

But here’s also (yet another) kicker: "Sell Fast, Sell Often" is NOT for everyone.

If you're running a $10 billion multi-stage megafund, this article isn't for you. Actually, you should probably do the exact opposite.

Think about it: Sequoia has to deploy billions. A16z needs to put $100M+ chunks to work. For them, riding winners isn't just optimal - it's existential. They NEED their Airbnbs and Stripes to take more capital every round. They NEED to double down, triple down, quadruple down. How else are you going to deploy $10 billion? $50M seed rounds? The math doesn't math.

These megafunds are playing an entirely different game. They're not optimizing for IRR or even DPI - they're optimizing for AUM and legacy. They're building dynasties, not scrupulously returning funds within 10 years. They need to catch a generation-defining company once a year, and ride along with its founders. They can afford to wait 15 years for an exit because they're raising a new $8 billion fund every 3 years anyway and LPs beg them to take their money.

But if you're reading this, you're probably not them. You're probably:

• Running a sub-$100M fund (hi, fellow emerging manager!)

• Actually needing to return capital to raise your next fund

• Unable to lead Series B/C/D rounds with $100M checks

• Incentivized on your carry, not your management fees

• Fighting for allocation in hot deals, not having founders pitch YOU

• Struggling to get LPs’ attention, not having them commit blindly

For us mortals, the game is different. We can't afford to wait 15 years. We can't keep doubling down. We NEED liquidity events to prove ourselves, to return capital, to raise Fund II before our kids graduate college.

This is why "Sell Fast, Sell Often" is essentially class warfare in venture capital. It's a strategy for the 99% of VCs who aren't in the oligopoly. It's David's sling against Goliath's $50B AUM.

The Cultural Revolution We Need

The biggest barrier to "Sell Fast, Sell Often" isn't mathematical - it's psychological. We've been conditioned to see selling as weakness, as "paperhanding," as missing out on the big score.

But here's the thing: venture capital isn't supposed to be a casino. It's supposed to be a sophisticated asset class that generates risk-adjusted returns. And risk adjustment means... actually adjusting for risk.

Look at the legends of prudence: Warren Buffett has built Berkshire not by chasing every mania, but by quietly exiting when the odds turned. Benchmark didn’t hold their entire Uber position to the bitter end — they sold significant chunks along the way and locked in one of the best venture outcomes of all time.

Or take La Fontaine’s tortoise — The Fables are a nightly companion for my daugthers these days. The tortoise didn’t sprint, didn’t get seduced by speed, but reached the finish line first. Caution, discipline, persistence: not glamorous, but decisive.

That’s the cultural shift we need in VC. From diamond hands to steady hands. From moonshots-at-all-costs to measured liquidity. The heroes of the next decade won’t be the ones who rode unicorns into the ground, but the ones who, like Buffett, Benchmark, or the tortoise, understood that slow and steady — plus partial liquidity — wins the race.

The heroes of the next decade won't be the VCs who held Uber from seed to IPO (though respect to them). They'll be the ones who built antifragile portfolios that survived the liquidity winter, returned capital to LPs consistently, and lived to invest another day. Unsung, unsexy heroes maybe, but still.

The Playbook for the Prudent

So how do you implement this? Here's “a” tactical guide. Feel free to create yours:

1. Set mechanical triggers: Once the MOIC hits 3x, sell 15%. At 7-10x, sell another 15%. No emotions, no FOMO, just process.

2. Use the SPV strategy for outliers: When something looks like it could 10x from here, structure it smartly.

3. Make connections in the secondary world: investment banks, specialized advisors, syndicates, family offices — the “secondary” part of the industry is increasingly structured. Know your way around it: those are your buyers and facilitators for such a strategy.

4. Track your alternative histories: For every partial sale, track what would have happened if you'd held. Learn from both your wins and losses.

5. Educate your LPs: Show them the math. Show them the improved risk metrics. Show them the 2021 vintage funds that are now underwater.

6. Frame it correctly: You're not "selling early" - you're "systematically de-risking while maintaining upside exposure."

The Punchline

Tyler Durden said "The things you own end up owning you." In venture, the positions you hold end up holding you - hostage to market cycles, liquidity events, and the whims of public markets.

"Sell Fast, Sell Often" isn't about giving up on greatness. It's about acknowledging that in a world where companies can reach $100m+ ARR under 12 months, where SPACs went from the future to the past faster than you can say “Chamath” and where "unicorn" went from rare to common to "please make it stop" - survival beats optimization.

The ZIRP era is over. The age of infinite liquidity has ended. We're now in the age of discipline, of distributions, of actually returning capital. Fundraising in 2025–26 will be brutal. The managers who embrace liquidity discipline will live to raise another fund. Those who don’t… won’t

So the next time someone tells you "only losers sell," remind them that the biggest losers are the ones still holding WeWork.

Sell fast. Sell often. Return capital to LPs. Raise again. Sleep better.

Enjoyed this contrarian take? Subscribe to Chasing Paper for more spicy perspectives on venture, finance, and the future of money.

Have a thesis on liquidity, secondaries, or why I'm completely wrong? Drop me a line, and always happy to debate over overpriced coffee in Paris or elseswhere.

• they were faster, agile and customer-centric: nimble and quick in product development through a novel methodology infused with MVP, design thinking and product management principles against more stage-gated V-shaped development processes

• they benefited from scalability: through the technology they leveraged, they could materialize increasing returns to scale (where a marginal unit can yield extra positive margin indefinitely) – and be appealing to venture capitalists looking for returns.

The combination of those features made them fierce adversaries for incumbents - old behemoths comfortably sat in their value chain - who feared “disruption” or “uberization” (not always with a firm grasp of what those concepts strategically entail, btw).

It was a David v. Goliath trope, a cautionary tale for large institutions to beware of fast-moving, customer-centric, tech-enabled competitors moving into their value chain and a beacon of hope for the small teams trying to come up with new approaches and business models in a range of industries.

So what happened in those 10-15 years?

Some startups have absolutely become inevitable and enacted the tale: Uber, Airbnb, Stripe, Revolut … set new standards in their respective industries and are now household names.

Many other cases are more nuanced:

• Slack was obliterated (at least in DAU figures) by Microsoft Teams, although it was an icon of product-led growth. Microsoft demonstrated that a traditional sales-led approach executed through powerful distribution and a complete product ecosystem prevailed. It joined Salesforce as a result to be part of a larger product ecosystem, but still lags behind.

• Disney+ showed it was possible for an establish incumbent to rival (again, in DAU metrics) with pure players such as Netflix or Amazon.

And from those nuanced cases, the David v. Goliath myth seems to fade.

Another remake ?

We’re now in 2025 and the landscape is even less decipherable.

It feels like we’ve crossed a threshold. Every company now has an “AI strategy,” just like they had a “mobile strategy” in 2010. OpenAI, Google, and Anthropic are releasing updates monthly, if not weekly. NVIDIA is worth more than the entire German stock market. Meanwhile, early-stage startups are pitching “GPT wrappers” at $30m valuations. This post isn’t about one specific moment, but the accumulation of signals. We’re entering a new equilibrium, and the startup playbook of the 2010s may not apply the same way in 2025.

Magnificent 7 companies (Amazon, Apple, Meta, Microsoft, Alphabet, NVIDIA & Tesla) have leveraged their pool of capital to trust the core layers of the AI revolution: foundational models, data centers and infrastructure.

• There is plenty of competition at the model layer, with constant leapfrogging for state-of-the-art capabilities. At the moment we have seen no evidence of a player figuring out continuous self-improvement of their model and take off vs. competitors, thereby dictating price. Quite to the contrary, the model layer is a knife-fight, with price per token decreasing fast for all models, including GPT-4 and likely to be commoditized going forward.

• Cloud services are the bread & butter of those hyperscalers. They use this cash to subsidize the AI value chain CapEx with their balance sheet, acting as risk-absorbers within the system, taking on as much demand risk as they possibly can, and driving the supply chain toward greater and greater CapEx escalation (a jargony way to say they invest in infrastructure and are set to absorb computing demand crunches for AI by deploying the capabilities).

• The next phase of AI development also emphasizes scaling physical infrastructure, with data centers becoming the critical battleground. Model sizes are set to grow 10x, requiring purpose-built facilities and optimal construction efficiency to stay competitive, driven by hyperscale CapEx from major players like Microsoft, Amazon, and Google.

All in all, the Mag7 are banking on AI and are positioning to reap in the profits to come. “Goliath” in this age is not a sleepy incumbent that has accumulated decades of fat and is clueless about the next technology wave: the hyperscalers are ahead of the curve.

But it’s not just the hyperscalers in the incumbent category: you have to toss in a number of now-mature tech companies trading in public markets (e.g Salesforce, Palantir, Intuit, Oracle, Snowflake, Datadog, etc.) or partially owned by late-stage PE firms (e.g Carta, Anaplan, Contentsquare, Highspot, Mirakl, etc.). Regardless of their individual trajectory and regarded as a group, those companies are no longer new contenders on the market, but neither are they lazy, slow-to-react incumbents. They all have plans to infuse AI in their products and their processes, product and dev teams (helped by AI) to deliver, distribution channels locked in and capital to make the necessary hires or acquisitions.

They’re a new character in our original story, a more advanced version of David but not yet Goliath: the tech incumbents.

And then there are the newly founded companies, the actual startups. Those that go on to raise pre-seed rounds at $20-30m valuation because YC set it as the norm. Those that brand themselves as “AI-native” and spin it as the advantage. Those that tackle often already addressed verticals with a new AI blueprint.

As an investor, I have to ask: what’s their right-to-win in this day and age? How can they beat incumbents with larger teams, larger datasets, established distributions, more funding and likely more talent? And can they ever rival with the hyperscalers or is that a fight never to be picked up?

Florent Quintini at Hexa has suggested a few ways AI-native startups can race ahead in their fight with incumbents: by reinventing workflows, capturing data more efficiently or accessing more data, by not being burdened with features rendered obsolete by AI, by transforming the business model or the distribution.

And he does make good points, serving untapped ends of the market (disruption) with new business models has been part of the startup strategic playbook for years.

One could also argue startups still have a few aces: velocity, focus, and permissionless thinking. They can chase edge cases incumbents ignore, break UX conventions, or test pricing models that feel “off-brand” for bigger players. They don’t have to maintain legacy features or appease large customers. That’s not nothing. But in a world where everyone can ship features fast thanks to GenAI and open infra, velocity may no longer be a moat - just a baseline.

And the question becomes, coming from a culture of product development that’s iterative and swift, how fast will incumbents adapt? How easy will it be to capture market share against competitors who not only have a headstart in distribution but also can deploy similar features rather easily?

If data and distribution are already claimed, what’s left for startups to defend? A few hypotheses, partly borrowed to Florent’s article:

• Emotional resonance: Crafting products people love, not just use.

• Workflow depth: Redesigning the entire workflow, not just bolting AI onto existing UI.

• Synthetic data generation: Building proprietary data loops through usage.

• Embedded distribution: Becoming part of the stack (API-first, infra-level hooks, product-led growth). Think Cursor

• Regulatory arbitrage: Moving faster in sectors where incumbents are hamstrung

These aren’t silver bullets, but they might be slingshots in the new David vs. Goliath setup. And I’m sure the final result will be nuanced once we post-rationalize it, but it will sure be fascinating to be on the battlegrounds with the founders building for a win in their markets!

Drop me a line if you’re a risk-taker running against established competition with a vertical AI startup - younes@oprtrs.club :)

And truly sorry for this post’s lame name – as promised last week, I’ll send these things raw.

May I suggest some Ol’ Dirty Bastard wisdom to get in the mood: ‘OOh baby I like it raw, yeah baby I like it raw’

Almost instantly, I had a few names in mind - musicians most of them: Charles Aznavour - who I admire so much he’s framed in my living room - Booba, Dr. Dre, Jul (sic). Other names came to my mind: Clint Eastwood, Martin Scorcese, Dave Chappelle, Cristiano Ronaldo, Kobe Bryant- and in the VC space, notably Fred Wilson.

And when I wanted to rationalize what all those people had in common, besides having produced some of my all-time favorite content, I realized that it came down to a very simple trait, yet a very hard one to master : they ship things consistently.

Indeed all of those people are absolute productivity beasts: Aznavour has recorded over 1,200 songs in 9 languages over a 70+ years career ; Booba has put out 12 albums and 5 mixtapes over 25 years (1 album every 1.47 years) ; Dr. Dre - who is famously slow at releasing his own albums - has still produced Snoop Dogg, Eminem, 50 Cent, The Game, Kendrick Lamar or Anderson.Paak in addition to building a headphones brand he sold to Apple for $3bn. Martin Scorcese has directed 25 films over 52 years (1 film every 2.08 years) while Clint Eastwood has directed 38 films over 48 years (1 film every 1.26 years). Chappelle is still on the road every single day practicing crowd work and writing new comedy lines in between tours and specials. Cristiano Ronaldo and Kobe Bryant (RIP) are known for their “hardest worker in the locker room” ethics, which their stats revealed. And of course Jul, with 24 studio albums and 10 mixtapes in 11 years, or an astonishing 3.09 albums per year (o/w 42x platinum, 3x diamond).

Of course, not everything these people ship is of equal quality - but they consistently practice their craft and make it a habit of putting out new work. While doing this, they increase their probability of having a hit or making a success.

And that’s the thing with streaks: the magic isn’t just in showing up - it’s in what showing up does to you over time. The act of doing the thing regularly, even when you don’t feel like it, lowers the activation cost. It normalizes the creative process. Each output becomes less precious, less perfect, more real. Habits compound not just in volume, but in depth: each new song, sketch, post, or pitch stands on the shoulders of all the previous ones. Over time, the delta between those who ship and those who wait becomes immeasurable - not because the former are better, but because they’re in motion. And motion compounds: more practice x more shots taken = more hits and better ones too.

And then there’s one I left out: Fred Wilson. For 16 years, Fred posted a note every day on his legendary blog AVC.com. He even wrote about the power of streaks, here or there for instance, and some data nerd has even put together an analysis of why AVC is a goldmine for investors, from early mentions of bitcoin and ethereum in Fred’s daily posts to tech trends and VC dynamics. Writing and committing to posting every day led him to explore more topics, to get a deeper insight into things, and surely did not impair his investing capacity, rather the opposite.

If you’re a reader of Chasing Paper, you know that posts are usually infrequent. Come to think of it, it’s not that I lack topics to tackle, in fact I have an iPhone note full of topics I’d like to write about. The truth is, I have a hard time getting started. Not because I don’t care - but because I care too much. I want each post to be researched, referenced, edited, layered. And with everything else going on - work, raising a fund AND baby twins, sleep (or lack thereof) - it’s hard to justify spending 3–5 hours polishing a 13-minute read that maybe 247 people will skim while half-watching Netflix.

But lately I’ve been thinking: these are the days of vibe coding. Of frictionless creation. Of pushing before you’re ready. So I’m gonna try vibe posting. Not in a reckless “tweetstorm and ghost” or “PSG won the Champions League. Here’s what it taught me about B2B sales ” kind of way but in a “write more, think out loud, don’t overthink it” spirit. Getting thoughts off my head. As frequently as possible. Right here. It might be raw. It might be GenAI-assisted. It might not be what you’re used to. But it’ll be real. Buckle up!

Subscribe now

Share

Despite this article’s title making him sound like a Big Bang Theory secondary character, Revolut CEO & founder Nik Storonsky is a bold and ambitious man. Not only does he want to make his company the #1 global bank, he also has a “side gig” revolut-ionizing Venture Capital.

His VC firm, QuantumLight, recently secured a cool $250 million in funding, with a very intriguing thesis at its core: AI can systematize and enhance VC decision-making.

“Really, is AI coming for my job too?!” I shivered internally, before recalling Daddy Marc (the a8n in a16z) confidently declaring a few weeks ago that venture capital was immune to AI disruption.

So which is it really? What questions does the Storonsky experiment at QuantumLight raise, and what does it reveal about the future of VC?

Let there be QuantumLight

At its inception, QuantumLight was the offspring of Nik Storonsky’s frustration with the VC value proposition : herd mentality, “strong” beliefs gone with the wind and market turndowns are markers of the erratic, human process of raising VC money.

Human bias, Storonsky argues, is the weakest link in traditional VC decision-making, and he aims to eliminate it completely.

Three years after announcing QuantumLight, the fund boasts a final $250m in funding, including $60m from Storonsky himself who definitely puts his money where his mouth is and 17 investments already made, all of which were recommended by AI with no exceptions. The fund will back startups in any sector anywhere in the world, at Series B and C stage. 70% of the firm’s portfolio are US-based, across AI, Web3, Fintech, SaaS and marketplaces.

QuantumLight isn’t shy about what’s under the hood either: its proprietary AI model, Aleph, has ingested data from the entire universe of venture-backed companies since the ‘90s and plows through 10 billion data points across more than 700,000 VC-backed companies to reveal success patterns.

More interestingly, when pitted against actual past VC performance, Aleph yields a 2x better performance than the top quartile firms (performance defined as average investment multiple).

How does it work in practice? Aleph screens thousands of companies every day, processes this information in real time, spots trends and positive signals from objective data points (speed of capital raise, overlooked market signals) and makes investment recommendations. Moreover, it updates its own models based on the outcomes of past investments: the model is self-healing and ever-improving.

And when Aleph builds up conviction in a given company, QuantumLight moves fast: one or two meetings max with a founder before a decision is made. Due diligence becomes data-first, human interaction-second.

Take, for example, QuantumLight’s investment in Mesh, a web3 infrastructure company. Aleph identified subtle market signals and data-driven indicators of rapid growth potential. Did Mesh pitch QuantumLight traditionally? Yes, but it was Aleph’s automated assessment that greenlit the deal. Humans were involved, but primarily to validate the model’s findings, not to drive them.

Revolutionary in private markets? Maybe. But systematic, algorithm-driven investing isn’t entirely unprecedented.

The Renaissance of VC

Storonsky’s experiment has echoes of a storied name in investing: Renaissance Technologies. As their website crisply explains: “Renaissance Technologies is an investment management firm that employs mathematical and statistical methods in the design and execution of its investment programs.”

Their Medallion Fund, a quant-heavy, algorithm-driven hedge fund with internal capital only, famously averaged annual returns of 39.1% after fees (66% before fees) over 30 years from 1988 to 2018. A $100 investment in Medallion in 1988 would have turned into a $400m fortune by 2018, and probably much more by now, since Medallion still managed to yield a 30% return in 2024.

The magic? A proprietary algorithmic approach that systematically identifies and exploits market inefficiencies. The most data-driven approach to investment there is. Sounds familiar?

I have no insider information to tell whether Renaissance was an inspiration for QuantumLight, but no doubt that Storonsky’s past as a quant trader at Lehman Brothers and Credit Suisse has ingrained this data-driven approach to investment deep in his system. And that’s probably the reason why QuantumLight has a team of engineers, data scientists, quant traders rather than your usual growth equity VC.

The question now: Can VC, traditionally a qualitative, relationship-driven game, mimic hedge fund success? Renaissance thrived in the liquid public markets, rich with minute-by-minute data. VC, on the other hand, is marked by opacity, sparse data, and illiquid, high-risk investments. Can QuantumLight replicate Renaissance’s method effectively?

AI on this "side of the table"

To fully appreciate QuantumLight’s experiment, it's essential to grasp the current state of the VC landscape, and boy is this landscape brutal!

Multi-stage giants like a16z and Sequoia increasingly dominate with ever more capital and deal access. Mathematically, a handful of firms having a significant portion of the total VC capital and top tier investor teams will be bound to have a significant portion of the wins - pressure is on for second and third-tier firms.

Not only that, but those multi-stage giants are also out pre-empting rounds months or even weeks after a previous close. It’s a frantic environment where VCs must decide often within weeks of a seed round whether to preempt with inflated valuations or risk being left out altogether. The competition is intense, and this intense competition requires VCs to have pre-existing relationships with founders, continuously track progress, and respond quickly to traction signals - "If it's Monday and you're hearing it for the first time and you gotta invest by Wednesday, you're just way behind".

Meanwhile, emerging managers face brutal fundraising climates, lacking the brand equity and access enjoyed by established players. Their only hope is to hop on successful deals very early and/or create a truly differentiated thesis and edge among this fierce competition to convince winning founders to go with them.

In a nutshell, at every stage, if you don’t spot and access winners early, you don’t win at all.

AI does offer an unfair advantage in this context: speed, scale, and the analytical capacity to detect and forecast early market trends from weak signals. AI models like QuantumLight’s Aleph quickly crunch vast datasets, identifying growth signals far ahead of human analysts skimming through LinkedIn for “working on something new” signals.

Additionally, operational tasks - from meeting scheduling to memo writing to due diligence to post-investment reporting - can benefit significantly from AI-driven automation: AI can easily 10x a VC’s output when used as a true Operating System.

Yet, limitations remain. Historical data might blind algorithms to unprecedented shifts, and qualitative factors - such as founder charisma, emotional resilience, or team cohesion - remain notoriously difficult to quantify algorithmically (although some founders I meet tell me that Hume AI is getting very good at perceiving emotions).

QuantumLight’s reliance on extensive data sets begs the question: from where and whom do these vast datasets originate? And once every VC has built their internal dataset, how do they turn them into an enduring moat like Medallion’s algorithm is to Renaissance?

Humans after all?

My take on the matter is that despite AI’s potential - that definitely needs to be leveraged by VC firms if they want to be successful in the coming years - there remains an undeniable necessity for human judgment in VC, especially at seed and pre-seed stages.

Data scarcity in early-stage companies forces VCs to rely on intuition, gut feel, and personal assessments of founders. Founder relationships are equally critical.

AI can’t meet you for coffee a couple of times when you’re iterating on an idea, throwing ideas. It can’t learn about your unique background while befriending you and getting to know you on a personal level.

And once AI has invested, although ChatGPT proves to be a great therapist, it can’t provide the emotional support a human can, namely the proverbial shoulder to cry on. Or the actual one: I vividly remember the day one of the founders of my firm’s portfolio found out that the lead investor in his current round was dropping. The world collapsed around him - but I happened to be there, and did provide the emotional support. Actually, what it took from me on that night was to order duck burgers, crack open a couple of cold ones and blast some Guns N’Roses out loud - but y’know. Fast forward to 2025, that company is profitable, growing, hasn’t raised a dime since and that founder is one of my best friends.

QuantumLight itself implicitly acknowledges this human necessity, recently releasing playbooks designed explicitly for coaching founders to high performance based on Revolut’s internal playbooks. Mentorship, emotional intelligence, strategic insights - these inherently human elements still play central roles, even if QuantumLight wants to systematically optimize them in their portfolio company.

As Marc Andreessen emphasized recently, venture capital remains a fundamentally human endeavor, even as the tools evolve.

And yet, it (re)turns!

The very nature of VC makes it too early to say whether Storonsky’s experiment will prove successful, but very simple to analyze: whether or not the fund has returns that are in the top decile of its vintage will be a straightforward, unidimensional way to tell.

Yet, incidentally, it’s the impact its very existence can have on the operations of VCs in the future that will be more interesting than the actual returns.

• Will all firms shift to AI-first investing and hire armies of engineers and quant traders or be subject to irrelevance? And if so, are we going to see ever larger multi-stage giants and watch all small-to-mid funds crumble into void?

• Will the human-first narrative gain a foothold and even be a differentiator towards founders? Like some sort of artisanal craftsmanship against the large industrial players.

We might be on the brink of an heliocentric revolution, the Galilelo moment of VC - when we collectively realize that the entrepreneurial universe doesn’t revolve around us investors, and maybe it should have been the other way around all along.

The comforting narrative suggests a hybrid future: AI augmenting human intuition rather than replacing it entirely.

But a more likely scenario? Post-traction investing could become overwhelmingly automated. Algorithms will analyze billions of structured and unstructured data points, including - even disturbingly - personal and behavioral indicators, from pitch dynamics to emotional responses, perhaps even genetic profiles (if we’re being dystopian).
Human investors morph into founder coaches — and it’s probably better if they have been through the circuit as founders themselves before doing so.

Pre-product market fit (PMF) investment, however, might always require human intuition. The maieutic act of uncovering, or even instilling entrepreneurial qualities in an individual, cannot be boiled down to data points. A number of past successful investments seemed irrational at the time and eventually succeeded spectacularly (e.g., Airbnb’s famously ridiculed concept at its inception).

To quote 50 Cent:

“I'm the diamond in the dirt that ain't been found. I'm the underground king and I ain't been crowned.”

Bazinga!

Traditional VCs, therefore, face a strategic crossroads: invest heavily in proprietary AI tools or risk obsolescence. Expect partnerships, acquisitions, and internal developments of AI tools to surge among leading venture firms.

Storonsky’s QuantumLight raises pivotal questions about the integration of AI in VC. The potential is evident: efficiency, speed, and perhaps superior returns.

Will AI be a mere augmentation or a complete revolution in venture investing? Can algorithms capture the intangible qualities of truly groundbreaking entrepreneurs?

Storonsky’s experiment is underway. How the VC industry responds to these evolving dynamics remains one of the most intriguing plotlines of our financial Big Bang Theory, but we do wish him and the firm a resounding “Bazinga!” at the end.

Share

“It’s over. It’s so over.”

As I yet again scrolled through my X feed to the tune of the vibecoders’ horde announcing the demise of tech as we know it and chanting the opportunity of making “$10k/month overnight without writing a single line of code - and buy my course to learn how to”, I proverbially screamed at my screen: ENOUGH!

Software engineering isn't dead; it's evolving.

In today’s startup ecosystem - part hype cycle, part existential crisis, part meme economy, part WTF is Trump doing - products like Lovable, Bolt, Replit, Cursor or Windsurf are populating the GenAI landscape and grab attention because of the insane levels of ARR they have been able to reach in a very short time (Lovable went from 0 to $10M ARR in 2 months). And yes, that is very impressive.

However, behind those storylines - I think it’s the deeper, strategic shift that deserves our full attention: what GenAI means for coding and the tech ecosystem in general.

Windsurfin’: a world of permissionless creation and permissionless distribution

If everybody had Notion across the USA, then everybody be vibecoding like Californ.ai / You’d see ‘em prompting their ideas and cloning repos too: a pushy, pushy software though. Windsurfin’ USA

Discarded draft, Beach Boys

GenAI has democratized software creation. Today, anyone equipped with a few prompts, ChatGPT and Lovable/Bolt can conjure up functioning software within minutes. To be clear: this is a profoundly positive shift, and the natural evolution of something that started decades ago with the advent of the Internet and has matured through years of Google, Wikipedia, YouTube, Coursera and TikTok: facilitated access to creation. To borrow an idea that’s dear to my friend Hugo Amsellem, it’s permissionless creation.

In the past decade, launching a fully functional business with an e-commerce website in a day through Wix, Shopify and Stripe became the new normal. These days, building an app or a SaaS can be done in under a day through Gen AI tools. This is the new normal.

A revolution? Maybe not, arguably the continuation of the same permissionless creation movement, nibbling on what seemed to be not gatekept, but at least only accessible to those who went through the knowledge and learning process of becoming a coder.

What GenAI allows - or seems to allow (more on that later) - is for the knowledge barrier to disappear.

Yet, this new abundance in permissionless creation inevitably demands equally permissionless distribution. Paradoxically, as barriers to software creation vanish, reaching and captivating an audience or clients becomes increasingly expensive. Customer Acquisition Costs (CAC) surge as noise multiplies and competition intensifies.

More competitors buy more ads targeting the same people in the same niche. More creators push more content with even less equipment and time spent producing. Hell, some of them aren’t even real people: they’re AI generated avatars mimicking content with the same positioning that went viral on TikTok (see Arcads, for instance). Even if you’re doing community-led growth, open-source, DevRel through X/LinkedIn and manage a lower CAC, channels are saturated.

We’ve entered an era akin to streaming platforms: endless content is produced, but discovering quality is harder than ever - and producing quality becomes an entirely new process.

So sure, you can expect to see hobbyists flourish and perhaps even marvel at a solo SaaS founder hitting a billion-dollar ARR. Yet these intriguing outliers won't define the industry's future or reshape markets at scale. Mass adoption requires capital, scale, distribution strategies - and yes, I’ll say it: proprietary intellectual property.

Enduring Moats: Proprietary Tech is King

I used to teach business strategy and tech at SciencesPo Paris. Invariably, I’d start introducing the students to the wonderful power and leverage of tech companies through an analysis of the production process:

In the industrial era, ideas need → capital to start → production. Products then need → distribution which is concentrated in a few retail stores to reach → clients.

With the Internet, ideas can be → produced for a few bucks online (provided you know how) and → distributed directly to → an audience through social media, newsletters, etc. Since production is vastly available and cheap to learn ; and distribution can be done directly, capital is no longer a barrier to reach an audience: it only serves as an echo chamber to scale. As a result, more ideas reach potential users, so it’s important to validate market-fit early before even producing or seeking capital.

Yes, yes … we’ve heard that story. Does that change with GenAI? No, it gets even more true as even the knowledge barrier falters.

However, because of the superabundance of permissionless distribution methods (influencers, newsletters, TikTok, LinkedIn - whatever you’re selling there’s a channel to reach your clients) acquisition becomes costly, and capital becomes crucial to scale. And it’s usually a VC’s role to fuel innovative tech ideas with capital.

Sure, you can also create a profitable operation and reinvest your profits in acquisition. In a world of superabundance though, what are the odds that a bigger, meaner, better-funded competitor reaches the market first with the same angle and captures the opportunity?

And what if you can’t consider that the very product that you’ve built is a moat because it was simply prompted into existence, which is replicable by anyone?

Here’s the rub: software built with GenAI tools, despite their disruptive potential, face a significant hurdle when it comes to venture investment: proprietary defensibility.

Serious investors generally don’t finance codebases easily replicable by another engineer with the same generative prompt. Proprietary technology, defensible data sets, and unique approaches are more crucial than ever to secure capital. Yes, cheap money has funded flimsy moats in bull cycles (and boy, is AI a hype theme!), yes some AI-wrappers have managed to raise funds but durable returns are likely to concentrate in proprietary data or infra.

Also, I’m talking about tech companies: software, AI, etc. – a marketing-first play, even fueled by tech, is different in essence. A funny example is OnlyFans ($6.6bn revenue in 2023, paying $1.3bn in dividends to its founder): while the site relies on tech that scales, the real core of the business is in how it partners with creators and helps them to monetize, not the tech itself. Back to our main theme.

I’m not saying parts of the codebase cannot be prompted: the same way some of it can rely on an open-source library, there’s no reason not to leverage this technology when it can be useful.

What I am saying is that ambitious tech projects need proprietary tech supervised by developers. Developers that can be augmented by AI, but definitely real humans that know how to read, write, build, comment, structure and correct code.

Tech teams: more crucial than ever

Let’s be crystal clear: the magic behind Generative AI isn’t in the black box - it’s in the humans who guide it. Knowledge remains the ultimate gatekeeper.

Without developers who understand guardrails, architectures, and legacy quirks, tomorrow’s systems will resemble spaghetti code drenched in AI-generated sauce impossible to untangle in 20 or 30 years. As nations and enterprises increasingly rely on software for defense, healthcare, and critical infrastructure, the demand for engineers with deep, vetted expertise is more important than ever, especially to build, guide and review those AI-generated systems.

This is why rock-star engineers are vaulting into roles that look more like hedge-fund gigs: AGI at OpenAI, rocket code at SpaceX, self-driving stacks at Tesla. Frontier engineers have a critical role to play in our collective future.

And yet, as recent data indicates, software developer job postings are down to 70% of what they were in 2020, and much much lower than peak demand in 2022. Indeed, we’re in a recession period and difficult access to capital + hiring freezes distort the numbers, but still: SWE are less in demand.

However, I see something interesting in it: it’s not software engineering that’s dying, it’s a certain sort of software engineer. What used to take entire squads of developers now gets shipped by a lone “product builder” wielding Copilot in a matter of days; sometimes hours. True tech companies have an output revenue per employee that’s higher than ever.

In 2025, the “middle-class engineer” is no longer a thing: the center has collapsed, the summit is reserved for a hyper-elite few armed with PhD or frontier knowledge, and everyone else is scrambling to be the AI-augmented one-(wo)man-show that can outpace a team. Is that so? Do we just write-off 90% of SWE jobs and call it a day? There’s more to it.

My friend Gilles Barbier has dubbed the mounting sense of “AI fatigue”: a weariness born not only of constant updates, endless model versions, and the ever-present fear of falling behind - but also from the perpetual need to review and clean up after an always productive, fast-shipping code tools spanning line upon line of code. Far from being plug-and-play, AI introduces layers of complexity - astute debugging, security concerns, compliance wrinkles, latency trade-offs - that demand seasoned navigators at the helm. This is not a bug; it’s a feature: complexity that can only be tamed by thoughtful, strategic minds.

Consequently, training and leading engineering teams will be more vital than ever. As AI systems get increasingly autonomous and code functions, then features, then entire products, our engineers will still need to “learn the craft” somehow. You can’t become an elite, frontier engineer overnight. You can’t become a senior developer able to supervise AI overnight. There’s no leapfrogging in human knowledge. And if we don’t want to create insane amounts of legacy, dependency and technical debt down the line, we still need software engineers that go through the learning process.

However yes, the developer journey is likely to be different. Entry-level “prompt coders” may pop up. Code curators, AI collaborators/trainers/reviewers, QA, AI-specific reliability, security or devops jobs will also probably be on the rise. However, the bulk of AI-driven innovation will hinge on those advanced engineers who can architect robust pipelines, version control data, and embed security by design. Also the talent geography can be impacted: remote AI collaboration tools can shift hiring to cheaper regions; as “middle-class engineers” may thrive in new markets rather than vanish. The software engineer of the future will be less focused on manual coding and more on system design, AI collaboration, integration, and oversight. New roles will emerge and career advancement will depend on the progressive mastering of AI-augmented development, strategic thinking, and the ability to orchestrate complex systems where human creativity and judgment remain essential.

It’s a bright future: one where the value of a dev increasingly lies not in their output, knowledge of many languages or tidiness, but in knowing how to identify the right problems to solve, knowing how to frame it correctly and leveraging an AI tool to shape a solution that’s robust, efficient, maintainable, scalable, understandable by other humans - and above all that does the job well.

And the coding itself? It will probably become a hobby for some - much like driving will be a hobby for circuits once countries start banning human drivers behind the wheel for safety reasons. Or it will be regarded as a curiosity from the past, like those accountants that used to calculate a P&L manually before Excel arrived (“oh wow, you did those ledgers BY HAND?! My God, I couldn’t do that”).

In short, commodity surface code is automatable ; systems thinking, architecture, and socio-technical stewardship are not.

From Enablement to Empowerment

Remember when Shopify emerged and empowered virtually anyone to become an online merchant? That seismic shift opened entrepreneurship to masses who never previously imagined running a business. Similarly, the GenAI coding revolution will open doors to an entirely new breed of software engineer-entrepreneurs: vibecoders, product builders, solopreneurs, creators.

Yet, just as Shopify didn't render e-commerce giants irrelevant, traditional software companies will persist. Handling, integrating, and modernizing tech infrastructure will remain essential: we have many more problems that software hasn’t tackled yet. Hell, “AI” is software itself.

Software development (in the sense, the development of software) is a cornerstone of future progress. Far from fading into irrelevance, software engineers will be more critical than ever to bridge the old and the new.

Software engineering is not dying: it's undergoing a metamorphosis. GenAI will democratize creativity and spark unprecedented opportunities, but the true industry shapers ; the tech companies that will tackle large-scale complex problems will master distribution, access to capital, proprietary IP, strategic enablement, and nuanced technical management. If you’re building one of those, we at OPRTRS invest in the founders that create those initiatives.

Far from the gloom of obsolescence, software engineers will find themselves central to a landscape richer, more complex, and more promising than ever. SWE's future isn't bleak - it's vibrant, sophisticated, and crucially, more human than we might realize.

As a quick reminder from the previous edition, in the past year, I’ve co-founded an investment firm targeted at operators: C-levels, VPs, etc. of tech companies who want to get together to invest in early-stage companies and roll up their sleeves to help them become future giants. If that sounds like you, do get in touch! 

As a result, I’d like to add new lines of content to Chasing Paper and to expand on investment theses for deals that we invest in. I’m a big believer of building in public for companies, as am I of investing in public. So here it goes.

I’m sure you noticed that AI was all the rage these days in VC-land and tech in general – and so I’d like to talk about something pivotal – search.  Specifically, the way web search, a massive ecosystem driven by advertising dollars (Alphabet, anyone?), is about to undergo radical transformations.

Shifts in Search

For decades, search has been fairly straightforward — Google ranks pages, and advertisers monetize them, bidding on keywords, acquiring traffic, etc. And for a long time, it worked perfectly, with a noticeable impact on all of our lives. Quoting directly from Alphabet’s 2023 Annual Report:

Larry and Sergey first wrote down our mission 25 years ago: to organize the world’s information and make it universally accessible and useful. They had an ambitious vision for a new kind of search engine to help people make sense of the waves of information moving online. 

It’s been inspiring to see what people have done with the answers to their questions, be it to find health care or comfort in difficult times, learn new skills, pursue new career paths, or start new businesses. The idea that a student in rural Indonesia could access the same information as a professor at Stanford was revolutionary and has changed lives and our world for the better. It’s opened up access to education and entrepreneurship like nothing else before it or since.

Google’s revenue from search ads hit $175 billion that same year, making up 57% of its aggregate revenue. But we’re at the start of something new: a seismic shift powered by AI and LLMs.

Whereas the traditional model feeds users results from crisp prompts, we’re now looking at systems that interpret queries in natural language and generate answers on the fly. And this matters because it transforms user experience from “best salmon recipes” or “how to calm crying twins” (oh yeah, this also happened last year) to receiving directly the top 3 best recipes, all integrated into one streamlined response, or, y’know, some methods that do not “actually” make babies stop crying but give you hope it can work. The AI doesn’t just show you where to find it; it serves it to you on a silver platter.

Obviously Google/Alphabet itself is at the forefront of that seismic shift, having gradually integrated AI into its products over the last decade and poured billions of dollars into cutting-edge AI research.

And yes, it is very unlikely that in order to get the same information you start typing “good day to you Ser GPT/Claude/Llama, could I interest you in providing me thorough details to turning a hen’s offspring into a deliciously seasoned pan-fried dish, colloquially known as sunny-side up egg? Thank you very much in advance, dear friend” instead of “omelet recipe”. What matters is that the format in which your result will be provided will be transformed. Right now, you’ll get a complete reply with the recipe you need from ChatGPT.

In time, you might receive an AI generated step-by-step video of how to actually make it, or an audio file, or whatever is most convenient for you to consume in your current setting, whether you’re on your computer, your phone, your VR glasses (have you seen those Meta Raybans?) or your connected fridge. Again, serving you the content you need on a silver platter.

A fundamental challenge here is that the web was designed for human users, not AI systems. Current AI approaches rely on mimicking human behavior - scraping pages designed for humans, complete with their visual layouts and ad placements. This inefficiency highlights the urgent need for AI-optimized frameworks that bypass traditional crawling methods, offering structured, machine-readable data instead.

But a more pressing question also arises: what happens to content providers? Those who make the recipes, blogposts, articles, reviews? Right now, the web runs on ad dollars: content creators rely on traffic to their websites, where ads generate revenue and/or content is paywalled. Every click is an opportunity for a site to display paid advertisements, with creators earning a portion of that spend, or to rake in payments or subscriptions. The more users visit, the higher the revenue: eyeballs have value. But if people aren’t clicking through to websites because they get their answers directly, how do creators make money? 

Hold that thought and save it for later, if you please.

Veridis Quo

The same shift, possibly bigger even, could happen to Discovery. 

A key change here is not just in how we search, but how we stumble upon things — whether it’s content, products, or services. Historically, platforms like Google and Amazon have perfected this balance through algorithms that mix content and ads so seamlessly, we hardly notice. Looking for “best headphones” on Google, get hit with results where you can click and buy directly. Search “headphones” on Amazon and receive most relevant products but also highest-ad-bidder’s product. 

Discovery is even more complex with social products, where your “feed” algorithmically pushes content that you might be interested in: Instagram, X, TikTok, etc. all use this mechanism to keep you hooked on their products, doomscrolling for hours and monetizing 

your attention all along.

You know the drill, we all do and have been used to it for decades now. With AI, things get messier—but potentially more exciting.

Product Discovery is already starting to change: TikTok is experimenting with automated product identification. Imagine uploading a picture of a pair of sneakers you saw a random person wearing in the streets, and AI finds the best deal for you for that pair immediately. Boom, conversion. No need to sift through links, search what brand/color/model those were.

Also notable, AI-driven search engines like Perplexity have begun integrating shopping features, allowing users to receive product recommendations directly within search results.

Content Discovery is also bound to change. If you type in on a LLM engine “what’s happening in the Middle East right now?”, an ideal result would yield a synthetic and balanced view of recent developments in the Gaza strip sourced through a number of articles from neutral and partisan journalists, local and international news outlets or official organizations - all narrated or put together by a single AI model. But you wouldn’t be able to access those articles, cross-check, fact-check, build your opinion on how one piece explains things more precisely than another, how one journalist writes in a style that is more enticing to you, etc.
To use a (slightly dystopian) culinary analogy; if all content is pre-hashed by LLMs, it might end up being like blending your entire meal before eating it. You get the same nutrients (sorta), but the taste is bland and you can’t identify the raw ingredients. 

Let’s leave it at that for the considerations of what the shift entails for individuals or for democracy, as albeit fundamental, it is not the topic of the article. Another aspect for which this is important is that when you perform search, you do not want low-value, generic replies as much as you do want the exact thing you are looking for.

For companies, it would mean that they need to rethink their entire sales and marketing strategies because their product could (or couldn’t) get “discovered” without traditional ads ; and for content creators it means finding new routes to monetizing what they do.

Advertise Aeternam

OpenAI, Perplexity, and Google are all racing to build the next-gen search experience, making it multimodal—essentially browsing the web on your behalf and giving you ready-made answers in palatable format. Great for users, right? But what about the creators? High-quality content takes real effort, and advertising is what keeps the lights on. It’s the engine behind news, social media, even paid streaming platforms.

Now, here’s the multi-billion-dollar question: if AI starts delivering answers directly, how does this impact the advertising machine? Google’s empire, fuelled by its advertising engine, is now under siege. If AI answers replace web traffic, who foots the bill for the creative effort content providers make? Where do the ads go? We could be heading toward an internet where quality content hides behind paywalls while AI regurgitates low-value, generic stuff for everyone else.

As AI replaces human browsing, the ad-based economic foundation of the web begins to crumble. AIs neither view ads nor interact with content as humans do, creating a vacuum in the revenue models that have sustained free content and services for decades. Without innovative monetization strategies, many content creators face existential threats.

Part of the solution might be that advertisers now start to pay directly for their products/contents to appear in AI-generated replies. And pretty surely the winner of the LLM race would love to rake in B2B ad revenue from placing ads in its replies, optimizing placement, algorithmically offering biddings, performance tracking, etc. But without safeguards and with the current state of AI (you know, hallucinations, etc.) this would mean pushing more dubious content into the cesspool.

Here’s where things get interesting for investors. If AI gets in the middle of this ecosystem, someone needs to pay content providers directly for them to keep providing quality content: scrapping the entire web without the suppliers making a dime is neither legal, nor sustainable from a value creation standpoint. Moreover, the web was built for humans, with UX and UI improving over time – not for robotic web agents crawling pages to retrieve information. Illegal, inefficient, costly, parsed: web access for AI is nowhere near where it needs to be for Search & Discovery to be displaced because of the economic equation in a world (potentially) without ads.

This is where Linkup comes into play. Linkup aims to build a tech-enabled infrastructure where AI can access content from providers, who get compensated directly, all in a format that’s adapted for a machine. It’s a transparent ecosystem that ensures creators aren’t squeezed out of the equation and get a revenue share in the value of the content they put in effort to provide.

Monetizing content through AI will surely require a fair and trusted intermediary that acts as a neutral third-party between content providers and LLMs.

This is why we decided to make Linkup our first investment; with a stellar team of ex-operators turned founders paving the way to the future of search & discovery in this AI-powered world.

Conclusion

The rise of AI demands a radical rethink of how the web operates : a new framework for browsing. Linkup envisions a future where APIs, licensing agreements, and AI-specific interaction layers replace the outdated scraping methods. These advancements will not only make AI browsing more efficient but will also provide transparency and fairness to content providers, ensuring everyone benefits from this shift.

Linkup’s model could be a key to preserving the economic equation of the web: content, eyeballs, value. The exciting part? This opens up new frontiers for advertising, where users are guided by AI but content creators, and brands get paid for the value they bring, traceably. 

It’s a more rational model that could prevent the “enshittification” of platforms / AI models, a term coined by Cory Doctorow to describe what happens when platforms prioritize ads over user experience.

So where does this leave us? In a space that’s exciting and uncertain. AI’s transformative power in search is inevitable, but we must shape it to ensure it benefits everyone — users, platforms, brands and content creators alike. As we edge into this future, companies like Linkup are the linchpins, offering a model where discovery, compensation, and transparency align.

In the end, AI won’t just change how we find answers. It’ll reshape how we discover everything – from your next favorite book to the new pair of sneakers you didn’t even know you needed, to maybe your next partner or who knows, how to put a baby to sleep? And for that, we need to change the Internet’s very architecture.

Until next time, Younes

If you liked this article and still aren’t a subscriber, you can join the community just here ; or like and share the article on social media 👇

Subscribe now

Share

In the past year, I’ve been busy working as a fractional CFO for a couple of startups or funds, but I’ve also co-founded an investment firm - OPRTRS (pronounced ‘operators’) - with my partners in crime Christophe & Noé.

If you know me (or if you read the last few Chasing Paper editions), you’ll know I love transparency, building in public and sharing the roadmap. So, in the spirit of Musk’s “secret” Tesla masterplan, I wanted to elaborate on the OPRTRS playbook and how we plan on redefining pre-seed investing. And sure, it’s not really secret – but after all, if Elon can manifest his vision by writing it, why shouldn’t I at least try?

Also, get ready for a lot of sports and Lord of the Rings references, because operating at the crossroads of an elite athlete team and a fellowship of determined friends just brought it out by itself (kind of).

The New (Pre-Seed) Deal

Let’s set the stage first. 

We are in France, and currently we see the French tech ecosystem entering its Third Age. 

But what happened in the First and Second Age? Well, in a hole in the ground, there lived a hobbit … sort of. When it comes to tech, an engineering-heavy education system has made France a stronghold for technical talent since the dotcom era ; however in the First Age that talent was lacking (i) sufficient venture capital allocations at the early and growth stages and (ii) an entrepreneurial mindset to go from zero to one, and one to a thousand.

The 2010s bridged the gap phenomenally, some successful entrepreneurs (X. Niel) and mafias (Exalead, Stupeflix, Fotolia) from the First Age paved the way to attract capital & business-savvy talent to the startup scene and to instill an entrepreneurial mindset onto French founders.

This led to a flock of successful businesses and pioneering ventures, among which: Algolia, Datadog, Dataiku, Alan, HuggingFace, Voodoo, Qonto, Zenly, Sorare, Aircall, Mirakl - and a new generation of entrepreneurial talent and angel capital ready to fuel them.

We see the Third Age tapping into those more recently acquired strengths and expanding on what sets France apart:

• deep technical expertise on frontier topics, including AI obviously but also bio-and-medtech, cybersecurity, web3, energy & climate, spacetech, robotics …

• a knack to build superior products: strong teams able to compete on product “plays” where product maturity needs to be very strong before selling, not only in consumer but also B2B (e.g : Upstream, Twenty, Jitter.video - that have the potential to be future Looms, Notions, Figmas …). This is especially relevant in a world where Product-Led Growth and Product-Led Sales are all the rage in go-to-market strategies.

• swift entrepreneurial execution and global ambitions from day 1: ambitious founders now understand that they need to expand to the US very fast (often at or before Series A) and know how to find advice and talent in the market to hone the internationalization playbook.

• the emergence of a top management layer brought by thriving scale-ups. We now have seasoned operators who are ex Apple, Stripe, Dropbox, or Spotify coming to work in French startups, and their playbook ripples through the ecosystem. We also have talents who have seen first-hand how to build and scale a tech company from France and onto the wide world. 

And the funding landscape? Well, historically VCs were the gatekeepers of early-stage funding, writing the bulk of checks, with a bunch of business angels who weren’t accustomed to the startup ways offering more than unfair conditions at pre-seed. But thankfully the landscape shifted, with more capital, local and international VCs specialized in pre-seed/seed and angels who provided fairer terms. 

Yet, the shift is not over: more super-angels and very-early-stage specialists are pooling capital, and guess what? They’re pulling off rounds without the need for traditional VCs. These angel syndicates are smart, they’re nimble, and they’re building a new normal at pre-seed: a world where founders raise directly from (i) risk-takers who know how to bet on a team very early without optimizing for ownership and/or (ii) the people who truly understand their industry and are ready to roll up their sleeves to help. No red tape, no fluff, just cash and know-how.

Brains Wanted

Enter the operator-investors. These folks? They’re the real MVPs of tech. 

Who are they? C-levels, VPs, Head of, Chiefs of Staff, General Managers, Country Managers, Directors, Leads, early employees … They’re the executive fringe of startups and scale-ups. Those are women and men that know their stuff, top talent who’ve got deep expertise, battle-tested scaling skills, and know the latest playbooks in-and-out. 

In the tech ecosystem, they’re invaluable. You don’t have to trust me, just see what those founders have to say about them:

"French Tech operators are the craftspeople of our ecosystem, those who know the ropes and will be the entrepreneurs of tomorrow. Without them, Tech loses its soul"
Charles Gorintin, Founder & CTO @ Alan

"Operators are those women & men working hard in the shadows of a cofounding team”
Anh-Tho Chuong, Founder & CEO @ Lago, ex-operator @ Qonto

"They are the pillars without whom the tech ecosystem would collapse”
Antoine Martin, CEO @ amo, ex-founder & CEO @ Zenly 

And indeed, entrepreneurs have started to realize the value that operators have, not only as their employees, but also as investors in their cap-tables. It’s increasingly common at pre-seed to have founders chase tickets from a handful of very capable operators from companies operating in the same space (for their industry knowledge) or with top-notch expertise in a key function for the venture they’re building (e.g a VP Sales for an Enterprise B2B SaaS which is by nature a sales-driven play).

Operator-investors are wanted and sought-after. But here’s the catch: their day jobs are intense, and there are only so many hours in a day.

The 110-meter Hurdles

Indeed, here’s where we hit a reality check. 

While operator-investors might just be the MVPs at the pre-seed stage, they still have got to clear a few major hurdles. It’s not a sprint; it’s a 110-meter hurdles race. If you’ve ever watched the Olympics, you know how it goes: the runners take off, looking slick and fast, but they’re bouncing over obstacles every 9 meters, and by the last hurdle, the legs are getting shaky. That’s what it’s like for operators trying to break into angel investing.

Let’s break down what’s standing in their way:

• Capital: While operators usually have (very) high salaries and sometimes even a cushy bonus, the truth is, they’re not exactly Scrooge McDuck swimming in pools of gold coins.
  Even if they’ve got stock-options in their company, their equity is often tied up and illiquid – and in the rare cases where they cash out, personal projects are obviously prioritized: wedding or baby on the way, world tour, buying property …
  As a result, most can’t throw four-or-five-figure checks at startups on a regular basis, even if they’d love to. They’ve got rent, maybe a mortgage, and definitely some healthy dose of Parisian croissants to fund. The idea of putting up 5-25k regularly just to get involved in deals is unrealistic for many of these high-potential but budget-conscious stars.
  

• Dealflow: The best opportunities don’t just come knocking. In the startup investment world, the juiciest deals go to those with an ultra-strong network. The irony? Our operators are so busy scaling companies, hiring, and fine-tuning their skill sets that they don’t have the time to network their way into every startup event. So, while they have insider knowledge and the brains to spot a future unicorn, they often miss the boat because they don’t see the deals early enough.
  And the double Kiss Cool effect (sorry for this very French reference) is that for lack of dealflow, even if they are willing to dabble in angel investing, they can’t diversify enough to moderate their risk.
  

• Time: Speaking of time, operators are stretched thin. Their days look like some complicated multi-level Jenga game where every block is another meeting, team review, one-on-one with a direct report or KPI to track. And when the day is over and there are no more meetings, that’s when the “real work” starts. Plus you know, personal life and everything. They’re not just “busy”; they’re “so-busy-they’ll-merge-with-their-laptops” kind of busy. Finding time to evaluate startup pitches? Most operators would sooner jump off a proverbial cliff than take on more responsibilities. They want to invest, but the commitment is daunting when they’re juggling a job, perhaps a family, and maybe a few extra side projects.
  

• Incentive Structure: The operator’s investment incentive structure is, frankly, lacking. Traditional angel investing can feel like running a marathon but being told you only get a medal if everyone else in the race gets one too. Operators have so much to bring to the table, but without meaningful carry or the right upside, why would they put in their valuable time? Investing a couple of small tickets, sweating it out with the founders to maybe get a serious multiple on one of the lines if they picked correctly (and that’s a big if). They might like the idea of angel investing, but without a real “win” in sight, the sparkle fades fast.
  

• Know-How: Lastly, angel investing isn’t something you can jump into without a plan. It’s a bit like cooking a soufflé: you need the right ingredients, tools, and timing, or you’ll just end up with a flat pancake of lost capital - believe me, I tried: so far I have 0 soufflés and not many more significant exits. Our operators have top-tier skills, yes, but angel investing is a craft of its own. They’ve mastered the scale-up game, but putting together a successful angel portfolio requires a different set of strategies.

So indeed there are important hurdles that make the leap from operator to operator-investor feel like a struggle. Sure, they’d make stellar angel investors, but most operators stay on the sidelines because of those painful elements.

OPRTRS: The Fellowship in full swing

Ta-da! This is where OPRTRS steps in, with a game plan to clear the hurdles and bring more operators to the investment world. On your marks, ready, set, let’s go!

We like to think of ourselves as a Fellowship of Operators. At the most fundamental level, OPRTRS is a community of top tech talent gathered by a will to share their expertise and skills with fledgling companies; as well as investing together.

Indeed, by gathering expert talent and pooling capital into a common vehicle, we can alleviate all the barriers standing in front of the operator-investors. We aim to empower the best operators in tech to finally become the investment force they were meant to be, and have fun while doing so.

What does it look like from the inside?

A. Simply the Best

We’re not just looking for people who can write a check; we want people who’ve been deep in the trenches - or should I say deep in the heart of Mordor. 

The C-levels, VPs, Heads of Product, and rockstar operators who don’t just understand the industry; they are the industry, and we want them to come from companies where they have been witnesses of grand successes. Below are some of the current or alumni companies where our 80+ operators to date have officiated.

But don’t think we just go logo-hunting. We’re after people with battle-tested experience and handpick our operators for the wealth of knowledge they have. They’re the ones who’ve taken products from 10 to 100, who know what it takes to scale, who’ve been in the war rooms, facing clients, upgrading interfaces, scaling the back-end, managing the workload, hiring the cogs in the machine. They’re not just smart; they’re in the know, which means they can bring real value to the startups we invest in.

B. We Alleviate Their Hurdles

• Capital: We don’t ask for big checks. Operators can come in with smaller tickets, and we match their investment with more traditional LPs who bring larger sums for leverage. As the great Muhammad Ali would put it: “float like a butterfly, sting like a bee” – get the best of brains and capital together and the impact is massive.
  

• Dealflow: By combining all of our extended networks, operators get to see premium dealflow they wouldn’t otherwise access. They’re very well positioned to know when one of their colleagues is going to leave and “work on something new”, way before VC analysts get the warning by a LinkedIn update. They might even get on to build something themselves. As a result, our community gets the chance to look at startups very early, across industries and that align with their expertise and interests, so they’re not just getting deals, they’re getting smart deals.

• Time: We know operators are strapped, so we handle the background work. From scouting to sourcing to vetting, our dedicated team takes on the heavy lifting. Operators can swoop in where it matters: engaging with founders, sharing insights, and making their presence felt without needing to be glued to a pitch deck.
  

• Incentives: Here’s the kicker: there’s real skin in the game. All of our operators chip into the investment vehicle, making it structurally in their interest to contribute to the success of our portfolio companies. Not only that, but operators who roll up their sleeves also get a generous slice of carry, turning their contribution into a genuine partnership. They’re not just “adding value”, they’re getting a return on the value they create.
  

• Know-How: Finally, they’re not stranded out there without a playbook. Our team consists of VC professionals guiding and handling the process. We have the tools, the resources, and the guidance to make solid, impactful investments and take care of everything in the back-office.

C. We Offer Clear Value to Founders

For founders, raising from OPRTRS isn’t just getting capital. It’s getting access to a strategic, power-packed community. OPRTRS doesn’t come in as a single check; we’re a constellation of expert operator-investors. I like to say that we’re a proxy investment for 100+ top minds, but you can think of it as getting the Avengers onboard your cap table, each bringing their unique powers to the mix (can’t use Gandalf & Legolas every paragraph, can I?).

Semi-decentralized VC

You often read about clubs or communities proudly stating that they’re essentially decentralizing Venture Capital. While to some extent making the claim and pushing an agenda towards its realization is progress on the market, there is still some advantages to having a centralized investment firm to deploy capital in startups. 

We see OPRTRS as getting the best of both worlds: the extended platform and the stable structure.

Indeed, we’re essentially structured around a community model that taps into the raw power of operator talent and networks and that makes us better at investing in early stage.

• Better Sourcing: With our network of operators, we don’t just wait for deals to fall in our laps; we’re actively out there, uncovering gems naturally. Our community-driven approach means we’re tapping into networks that traditional VCs only dream about, and we increase our chances of stumbling upon one of those very few founders each year that is going to build a generational company.

• Better Screening: The collective experience of our operator-investors means our screening process is rock-solid. We don’t just assess ideas and teams; we can drill down to assess technical edges, strategy, viability, scalability, and all of the intangibles only operators understand.

• Better Winning: Picking OPRTRS for a founder isn’t hard to do. It’s not a leap of faith to uncover later whether there is value-add or actual industry knowledge beyond capital. When a founder is picking OPRTRS, they’re choosing more than just capital. They’re choosing direct access to and a seal of approval from the industry’s top talent. This isn’t just money, it’s a strategic alliance.

• Better Supporting: Unlike many VCs, our post-investment support is in a league of its own. With a community of operators at the ready, founders get access to real-world expertise, on-demand, and from people who’ve done the job themselves.

That’s the decentralized part: our community doesn’t just make us different; it makes us better. But we also pride ourselves in having a stable operation by centralizing a lot of the process behind the scenes: there’s a team of seasoned investors who handle the heavy lifting. Think of us as a VC engine running in the background so operators can focus on delivering their expertise to our portfolio without getting bogged down.

We take care of everything: dealflow, due diligence, memo writing, fundraising, compliance, community management, operations to mobilize the right operator at the right time, legal, admin. Essentially, VC-as-a-Service. Our goal is to let operators zero in on their strengths: guiding startups, sharing their know-how, and driving real value, without sweating the back-office details.

With this model, we give operators the freedom to be as hands-on as they want with founders, while we handle the complex, time-consuming work to make investments and support seamless. It’s a setup that allows everyone to do what they’re best at, keeping the engine running smoothly and the impact high.

Building an Ecosystem Flywheel

What’s our secret you ask? 

Our vision isn’t just for now; it’s for the future. Today’s operators? They’re tomorrow’s entrepreneurs. OPRTRS is about building a sustainable ecosystem: we gather operators, have them connect, invest in their ventures, and support their growth. In turn, these founders hire and train new operators, who eventually become the next wave of founders and investors. This is the long game: it’s about building a regenerative ecosystem that feeds back into itself, creating a perpetual cycle of talent, capital, and innovation.

And just like that, the flywheel spins.

There you have it: the “secret” OPRTRS masterplan. It’s big, it’s ambitious, and it’s rooted in building something that lasts. So, if you’re an operator, a founder, or someone who believes in what we’re building, welcome to the future of pre-seed investing. This is only the beginning.

Until next time,

Younès

If you made it down here, kudos! For those of you that don’t know, Chasing Paper is a newsletter giving my perspectives on finance, VC, strategy, tech & innovation - and how they interact. I haven’t been posting in a while as I was busy building but I plan to post more in the foreseeable future.

If you’re not a subscriber 👇

Subscribe now

I recently left my position as CFO & Director at the investment firm The Family, and am taking the time-off opportunity to write about topics I haven’t talked about while figuring out my next full-time gig.

As a refresher, Chasing Paper is a newsletter giving my perspectives on finance, strategy, tech & innovation - and how they interact.

I previously shared 10 Thoughts on the Gaming Industry. Today, I’m doing the same with Luxury as a target industry. With Bernard Arnault (Chairman and CEO of LVMH) recently crossing the mark to become the richest person in the world, diving into the luxury goods market is a source of valuable insights on how to hold steady in crisis time.

Warning: this is a long post.

Enjoy, and subscribe if you haven’t!

Subscribe now

My interest in the luxury industry has grown over the years. While as a teenager I was a mere aspiring customer, I realized in grad school that the prospect of working in the industry was tantalizing to many and started wrapping my head around it by reading CEO profiles. 

With the years passing and my disposable income growing, I started to enjoy luxury experiences (for me, it’s fine dining) and reflecting more and more on the drivers and dynamics of this one-of-a-kind industry.

More recently, my wife-to-be became a concierge, working in one of the most exclusive hotels of the world. That was another insight into the industry, from the employee perspective this time.

The 10 thoughts below are personal views developed over the years.

I couldn’t think of anything screaming luxury more than James Bond in a neat suit in front of an immaculate racy sports car

Clientele

1/ Shopping for Status 

Growing up, I was always a massive fan of Italian supercars: marvels of automotive design, with incredible performances, sleek curves and bold colors. I regarded the ownership of one as a token of life achievement: ‘when I’m rich, I’ll get an apple-green Lamborghini’ (sic).

While I did snap out of it - admittedly not entirely - for a number of reasons (including but not limited to: learning about asset depreciation, living in a car-banned city-center, crypto bros, Andrew Tate), I think it’s a relatable experience and a good introduction to the main driver of the luxury market: desirability.

Whether your kick is clothing, jewels, bags, watches, spirits, dining, hotel leisure, cars, boats, planes, art, real estate: the luxury market has something for everyone to crave and lust for.

Arguably one of the best articles I’ve ever read about tech is Eugene Wei’s Status as a Service (do yourself a favor and read it). Eugene’s article also offers incredible insight into human nature:

“Let's begin with two principles:

• People are status-seeking monkeys*

• People seek out the most efficient path to maximizing social capital”

The analysis that follows is breathtaking, unraveling how social capital (where we stand on the status ladder) is much harder to grasp than financial capital, yet how the former serves as a proxy indicator of the latter. Ergo, the more social capital you have, the likelier you are to own high financial capital.

My first observation, as for the luxury industry, is that many believe that the opposite is true. While people seek status, displaying financial capital appears as the most efficient path to maximizing their social capital. 

And what better way to display financial capital than to carry a Chanel bag, wear a Patek Philippe, drive a Bentley, smoke a Cohiba, own a Mondrian or live in a secluded mansion? 

Of course, the argument can be made that those are all fine pursuits for genuine amateurs: if you truly enjoy leatherwork, watchmaking, automotive engineering, cigars, art or real estate (respectively) then you would aim for artifacts like these. The point is that the two are non-exclusive: true amateurs and ‘wannabes’ alike display financial capital when purchasing luxury goods.

As such, luxury goods customers are shopping for status, and luxury industry players are purveyors of one of the most sought after resources on Earth: status. Better yet, readily-available status. 

That fact alone explains why luxury is such a good business, especially in times where status (whether in the form of likes, followers, retweets, views, dollars or votes) are increasingly considered important.

2/ A bottomless market 

Another reason that makes the bull case for the luxury industry so compelling is that it is a virtually bottomless market.

I actually realized this upon reading XAnge’s Cyril Bertrand excellent article on his investment thesis upon investing in LeCollectionist back in 2017.

Cyril says:

“I often have to remind myself that luxury is not a niche. True luxury — as a market, not as a branding strategy — does not show off on the Champs Elysées. It is a discrete, demanding, disciplined segment, much, much deeper than we spontaneously think.”

Inarguably, luxury is a spectrum: there is little to no commonalities between the teenage US buyer of the lowest priced Balenciaga bag and a South Asian tycoon lavishly spending $100k+ on hotel stays on a regular basis. 

Arguably, however, both ends of the spectrum add to the industry profitability: low(er) margin, high volume luxury goods at one end ; very-high margin, low volume at the other - but more on that later. The industry indeed serves a large type of customers, from the highest brackets of middle classes (HENRY: high-earning, not rich yet) to HNWI and UHNWI ([ultra] high net worth individuals).

In all cases, though, the pricing elasticity is huge: buying appetite is of course a function of price, but with a <1 coefficient (i.e the breaking point for a single purchase is reached slower as price increases). 

Considering the range of the consumer base, their very elastic buying appetite, the virtually infinite buying power at the richer end of the spectrum, the regions who just “recently” started to consume luxury goods and crave for it (esp. China, Middle East, Russia), and the number of different categories (clothing, jewels, watches, spirits, dining, hotels, cars, yachts, PJs, art, real estate) – the luxury industry is indeed seated on a growing pot of gold, as even the most niche products can very well tap into immense addressable markets.

3/ Now you see me, now you don’t

Something quite puzzling about the luxury market is how ubiquitous it is - with ad billboards and shops in city centers, malls, social media, airports, etc. - and yet remains exclusive and very high-end.

My take is that most of the market is actually hidden in plain sight, with the more confidential and exclusive experiences accessible only to the highest spending consumer segments and the entry level made as visible as possible.

Indeed: you will see prospective customers line up on the Champs-Élysées (likely tourists) in front of the Louis Vuitton boutique to buy a bag or a pair of shoes (which is a droplet in the brand’s sales: LV just had a booming year, topping €20bn in revenue), while more elite purchasers spend (much) more but have the boutique privatized and/or a personal shopper and/or have items delivered directly to their hotel or pied-à-terre.

We mentioned just earlier the different luxury consumer types: from high-earning upper middle class individuals (HENRYs) to really ultra high net worth individuals (UHNWIs). Inarguably, the marketing strategies that brands need to deploy to target those different segments are in essence very distinct, as are the revenue and profitability breakdowns for those market segments. 

Some brands, however, are cleverly playing across categories. Paris Society is a French hospitality group offering some of the most currently sought-after restaurants, clubs and events in France (Paris, the Riviera and the Alps). Recently acquired for a whopping €330m by the Accor hotel group, Paris Society has built a recipe that’s quite unlike the others in the market: exceptional and iconic venues that appeal to locals and tourists alike, quality cuisine but no emphasis on sophisticated menus and dishes, and make everything Instagrammable.

Interestingly, all categories of luxury consumers are heavily satisfied with the concept: very-high spenders see it as casual but chic, far from the bling they’re used to (“understated sophistication”), while the venues are aspirational for many entry-level consumers. The concept will be rolled out internationally thanks to Accor’s international leverage, and surely will find resonance.

Business dynamics

4/ A European exception?

Can you think of five luxury brands? Okay? Go!

Out of those five, how many were not European? Maybe you thought of Tiffany’s, Marc Jacobs, Fenty Beauty, Supreme (🇺🇸), Comme des Garçons, Yohji Yamamoto (🇯🇵), or even De Beers (🇿🇦).

More likely, your suggestions were French 🇫🇷 (Dior, Chanel, Yves Saint-Laurent, Hermès, Balmain, Céline, Louis Vuitton, Givenchy, Louboutin, Lanvin, L’Oréal), Italian 🇮🇹 (Gucci, Prada, Versace, Bulgari, Ferrari, Lamborghini, Dolce & Gabbana, Valentino, Armani, Loro Piana, Off-White), British 🇬🇧(Bentley, Rolls-Royce, Stella McCartney, Vivienne Westwood, Burberry, Alexander McQueen), Swiss 🇨🇭(Rolex, Tag Heuer, Hublot, Patek Philippe, Audemars Piguet, Vacheron Constantin), German 🇩🇪(Rimowa, Porsche, Mercedes, Maybach), or Spanish 🇪🇸 (Balenciaga, Loewe, Massimo Dutti, Manolo Blahnik).

As a long observer of the European tech market, and European industries in general, I have to admit that there are many industries were local European players appear to compete on the sidelines, while the US and Asian competitors fight for the lion’s share.

The luxury market is very distinct to that regards, with the top brands mostly being originated in Europe, as is also the case for the leading conglomerates (LVMH, Kering, Richemont, EssilorLuxottica, L’Oréal, Hermès - with the notable exceptions of Estée Lauder and Chow Tai Fook).

Lake Como, a European landmark and go-to destination for worldwide luxury consumers

The prevalence of European luxury brands, including the more recent ones (Off-White, Margiela) stems from historical consumer perception that these countries produce luxury goods of superior quality and have precision craftsmanship. Indeed, having been involved earlier in leatherwork, watchmaking, engine building ; having developed hospitality arts, and having colonized a sizeable portion of the world to secure access to the most sought-after resources, European countries had a head start in the luxury market, that they retained through efficient industrialization and transfer of skills, but also a lot of marketing: it’s not as good if it’s not from Europe (which by the way makes for hilarious memes 👇)

Champagne is a protected designation of origin and is only called champagne if it comes from the Champagne region of France, otherwise it has to be called sparkling wine, which infuriates many people on Twitter.

This concept, the provenance paradox, dictates that a product's country of origin establishes its authenticity and quality. Consumers associate certain geographies with the best products, such as French wine, Italian sports cars, and Swiss watches. Products from other countries, especially developing markets, are perceived as less authentic and even if they are of the same quality, less desirable. 

With desirability being of the utmost importance for luxury brands as we established, it’s no surprise that European brands and groups have retained their importance in the market. 

5/ A steadily lucrative market

If you’re any interested in annual financial reports (which, if you’re reading a newsletter initially designed for CFOs and finance professionals, I assume you are at least a little bit), then I suggest you have a look at LVMH’s 2022 financial performance document.

A caveat before diving in: obviously, the group has its own strengths and weaknesses, but its size and scope make it in my opinion a good enough proxy for the entire industry. However, things that are said below might not apply to independent brands, who will have different challenges (procurement, production, overheads, marketing, etc.).

In the report, several of the revenue and profit breakdowns are very interesting, including the ones by regions, currencies and categories. More interesting are the operating totals for the entire group: 68% gross margin and 26.6% operating margin on a €79.2bn revenue.

A few remarks here: 

• One could think that luxury goods, only sourcing and using quality materials (whether it’s leather, grapes, cashmere wool, tobacco leaves or precious stones) and requiring precise craftsmanship would have low gross margins. That’s not the case: gross margin stands very high (with high variance across categories to be noted). That’s also an advantage of conglomerates over small, independent brands.

• Conversely, one could think that luxury conglomerates got so rich by industrializing production processes and decreasing procurement costs while increasing retail prices. Again, not really the case, since production costs account for a significant portion (around ⅓) of revenue - factoring in raw materials, design and craft. The “you’re paying an x00% premium just because Gucci is written on the T-shirt” argument does not stand, overall.

• Looking at both points above, I realized that a long-held thought of mine about the luxury market is disputable. I always thought that there was an immense discrepancy between perceived quality and actual quality in luxury goods. As such, marketing stunts would create a desire bias for goods that overall aren’t that well produced. Conversely, at the other end of the spectrum, going the extra-mile to create a really unique experience for the most demanding customers would kill profit margins. While naturally we do not have access to an itemized profit breakdown by SKU, and I imagine well that each individual brand has its cash cow products and its lower margin, higher prestige ones ; the result is well-balanced at group level. Having a brand and product portfolio surely helps mitigating both effects.

• Final remark, marketing and selling expenses are the main cost element in the income statement, at ~35% of revenue. Designing and producing luxury goods is costly, but even costlier is the art of making them desirable, as we shall see in the next section.

6/ Stargazing distribution

There is too much to say about luxury marketing to fit in a single article. However, one aspect I find rather interesting is how the status appeal we mentioned in the first section is reinforced or even sometimes created by savvy marketing.

One example of that is endorsement deals: you have probably seen countless billboards with a singer or actress in a panoramic scenery trying to sell perfumes. 

Perfume licensing deals are indeed an alluring prospect for celebrities, who are typically awarded a royalty of 5-10% of the sales, along with mouthwatering upfront payments for lending their name. 

But there’s even more in it for the brands: celebrities are often involved in luxury marketing because of their popularity and influence over consumer behavior. People are drawn to their lifestyle and image, and associate that with the products they endorse. If you’re a status shop, displaying high-status individuals who endorse your products is quite obvious as a marketing strategy.

A few remarks:

• Some brands, Hermès for instance, are reluctant for their own reasons to rely on celebrity advertisement deals. Is that to say that they can spare celebrity endorsement or references? Well, not really. Hermès’ iconic bag is the Birkin, literally named after actress Jane Birkin. What’s more, even though it does not rely on KOLs (key opinion leaders) to sell its products, Hermès produces the exact same effect by having celebrities, millionaires and other high-status individuals wait and crave for their products. It makes their brand even more unique: wanted by those who have it all. As a side-note, Birkin bags have shown a steady 14% YoY increase in value over the past three decades, making them one of the best alternative asset class there is.

• Some other brands, mostly upmarket ones (Boss, Calvin Klein) but also some brands geared towards discount consumers (CVS, Walmart, etc.) also use the same branding tactics and obtain celebrity endorsements. The strategy is effective, still consumers know what brands are considered luxury due to craftsmanship, exclusivity, heritage, price and other elements.

• Luxury brands increasingly rely on influencers as the content creator space matures and allows them to turn eyeballs into dollars. For instance, Dior’s partnership with French influencer Léna Mahfouf had a media impact value (the sales impact of placement, endorsement and other social media campaigns for a single media) of $4.72m in 2021.

In any case, luxury brands are champions at creating masterful marketing buzz and receiving endorsements of household or up-and-coming celebrities is a definite trick up their sleeves to continue enticing consumers of all ages, genders and regions.

As Bernard Arnault put it himself: the luxury business is difficult because “you need to create desires for things that no one really needs”. Showcasing a galaxy of stars definitely helps.

7/ The talent war

One aspect we haven’t discussed yet is how all luxury brands, small and large, all emerge from a determined Creative and Artistic endeavor. 

In general, luxury brand identities are defined by their founders (think Chanel or Yves Saint-Laurent), who give it a clear DNA. From there on, the brand is put in motion and reinvents itself through the work of creative designers who embrace the brand history and are able to express it with their own eye and renew the brand expression.

It’s all about individual talent and risk-taking. Finding or hiring a Karl Lagerfeld (Chanel), John Galliano (Dior), a Virgil Abloh (Louis Vuitton) or an Olivier Rousteing (Balmain) is a quintessential task for all brands, who are also putting the crown jewel at risk every time a new Artistic Director is hired.

As with all markets where individual talent is in very high demand and very short supply, there is a continual mercato (transfer market) for talent. This is true for fashion, but also restaurants, events, and other forms of luxury goods and/or services.

Gary Cruz, Master Watchmaker for Audemars-Piguet

Yet, besides the essential role of Artistic Directors, Designers, Chefs and Architects are also countless jobs that the industry requires in order to maintain its prestige: haute couture workshop fairies, leatherworkers, shoemakers, watchmakers, craftswomen, concierges, tailors, sommeliers, maîtres d’hôtel, etc. Luxury groups invest a great deal of money into the hiring, training and retention of these qualified workers who are also essential to the quality of product/service.

It should also be noted that the industry also hires a great number of less qualified workers: kitchen helps, clerks, busboys, dishwashers, petites mains, etc. who do not benefit from the same prestige, yet also suffer the extremely demanding environmental pressure. As an employee in the industry, you are supposed to deliver the best only, which has many pitfalls. 

René Redzepi, the Chef of the now-closed-once-best-restaurant-in-the-world Noma, puts it this way: “fine dining, like diamonds, ballet and other elite pursuits, often has abuse built into it.”

8/ Conglomerates and independent brands

Caravaggio’s David & Goliath. The smaller one doesn’t always win, though.

M&A activity in the luxury space has grown continuously over the past few years.

Under increased competition, consumer behavior changes and the need for scale and efficiency, large conglomerates (Kering, LVMH, Richemont, Estée Lauder, L’Oréal, EssilorLuxottica, Chow Tai Fook, etc.) have grown ever bigger in the past years through a trailblazing acquisition spree.

Indeed, pressure is increasing for individual, independent brands: expanding their reach to new regions or customer segments, increasing market share or launching new product lines are all very costly endeavors in the presence of behemoths in the market. Consolidation has allowed luxury brands to scale and improve distribution, pool resources, share expertise and reduce costs, which improves competitiveness - and makes the acquirer ever profitable with a portfolio of balanced brands.

The question is: can individual brands resist the movement?

Some have: Hermès, Ferrari or Rolex remain independent to this day. Yet for them, the key to remaining independent lies in the mythical character that they have for consumers: legacy is a great help.

For newer brands, the question remains. There is no definitive answer to this question, but I believe in a few assets that independent brands have over large conglomerates that they can strategically leverage to their advantage:

• Creativity: as we mentioned in the previous section, creativity and talent are the centerpiece of luxury brands. Designers who build their own brands can be radically experimental and entirely expressive of their identity, which can gain foothold rapidly and win over some consumers.

• Risk-taking: similarly, independent brands can take more risks in their marketing approach, distribution techniques, etc. They do not have to be ubiquitous, rather they can try to gain momentum and market shares in specific, niche markets.

• Adaptability to trends: while consumer behaviors change (m-commerce, live purchases on TikTok, carbon neutrality, etc.), smaller brands have larger avenues to experiment and cater to these new consumption patterns

• Innovation: finally, independent brands can also incorporate tech and innovations more easily into their product offerings, whether it’s on the product itself (e.g 3D-printed dresses) or in the way it is marketed, distributed or sold (e.g virtual showrooms, VR catwalks, etc.).

While they don’t have the scale, resources or cost advantages, smaller brands can count on some strengths that are core competitive advantages in the luxury market: a flair for design, innovation, risk-taking and limitless creativity.

It will be very interesting to see how the market plays out in the next few years. 

My take is that a few brands might slip to the crack and make it to a worldwide scale without being absorbed by larger conglomerates, but that this will be the minority. Large luxury groups will continue to expand their portfolio of brands and grow externally through M&A, raking in more and more revenue and profits.

Conveniently, time for us to move to the perspectives for the luxury industry.

Perspectives

9/ Tech & Luxury : friends or foes? 

The luxury industry, like all others, has not been spared by the need to implement digital innovations into its operations. All core functions, product and supply chain management operations, distribution channels have digitized and will continue to overcome silos and challenges linked to digital transformation. 

That is not the point here: obviously being able to sell on mobile, through influencer social media, to model new products through digital tools, to adapt wrist bands to digital watches, etc. are things that luxury brands need to do in order to cater to changing consumer needs.

What I would like to talk about is that luxury and tech seem to have a love/hate relationship. They could pair very well, yet they tend to stay apart.

Indeed, luxury has an intrinsic attachment to tradition and legacy ; yet it is also the domain of pure creativity and innovation - making it an interesting object for tech.

For instance, there is no such thing as luxury software. Sure, there is expensive software (think anything B2B that very few large customers badly need), but there’s no status or desirability elements attached to it. No one will tell you “oh my god, I am so jealous of your new Tidal/Netflix/Slack subscription”. The one example I thought of was the infamous ‘I am Rich’ app that debuted with the first iPhone at a $999.9 price and that did nothing but signal you had bought it. Pure status.

Yet, as SaaS develops, you could think of alternative strategies to increase MRR whereby number of seats is limited and ARPU (average revenue per user) is much higher, and you rely on user FOMO to acquire more customers. Companies like Superhuman have used this strategy with great success when launching.

At my previous firm, we used to say that tech is the perfect tool to provide luxury-like experience at scale. Indeed, tech allows to personalize services or products offered to each individual users, based on feedback or usage data and/or selected preferences. But is it really luxury if it reaches scale? Doesn’t luxury imply some form of scarcity? And isn’t that intrinsically incompatible with technology revenue models (increasing returns to scale)?

Apple might be the one brand that has a foot in both markets. It builds technology products and markets them like luxury. Yet its products are hardware, physical ones. One could argue that iOS user experience is high-end, hardly that it is luxurious.

The Merge sold for $91.8m

One of the recent innovations that could reconcile the scarcity of luxury and the scalability of technology are NFTs (Non Fungible Tokens). In the past years, pictures of monkeys have sold for millions of dollars as if they were an original Van Gogh. While the raving about NFTs has virtually stopped with the slowdown of the overall cryptocurrency market, the concept itself could help implement luxury-like concepts into digital products.

Arguable, to say the least: while Alexandre Arnault (heir to LVMH) is the proud owner of a CryptoPunk, his father Bernard has taken a jab at competitor Gucci for selling virtual sneakers: “we’re in the real world selling real products”.

10/ The future of luxury

Wrapping up, I’d like to talk about the future of the industry - not by stating possible truths (my crystal ball is broken these days), but by raising a few key questions that will surely determine how the space will move going forward.

• Is status-shopping going anywhere?

Is it likely that in the next few years, the human desire for personal recognition and ego boost will fade away?

While my opinion is that society seems to be going in the opposite direction (with likes, views, clicks, etc. becoming a norm), we might reach times where more pressing needs in the Maslow pyramid need to be addressed first due to ecological or societal reasons.

However, I still think that even if constraints appear (consumption, energy, etc.), inequalities will rise and allow a smaller group of individuals to access more demanded luxury products (artificially scarcitized).

Overall, I think that looking for status is inherently human and that luxury is a mean to that end - an efficient one at that as it offers Status-as-a-Service.

• Can luxury goods become digital, and if so are luxury brands equipped to pre-empt the move?

I’ll leave that one as an open question - my thoughts are in the previous section.

• What effect will recession have on the luxury market?

This one is more current than the previous ones. While stock markets are plummeting and inflation is increasing, what can we expect from luxury consumers?

On one end of the spectrum, will consumers save more money? Channel their disposable income to more pressing needs? Or will they invest in alternative, “safe” investments (as we saw, luxury handbags are among these).

Overall, the pandemic has benefitted luxury brands and the rapid expansion of the customer base (especially driven by Asia and younger consumers) could be partially offset by a decline in revenue per consumer while recession lasts.

Yet, luxury conglomerates have had exceptional results in 2022, have profitability levels that make them geared to face the long winter and can count on customer segments that recession does not really affect.

Having written this long-a** article and raised those questions, I have to admit that I am very long luxury. Happy to have a chat about it if you disagree or would like to dive in a particular topic more specifically - as you’d have noted, my guilty pleasure is a gourmet meal, so you know where to invite me ;)

Until next time,

Younes

Long time, huh? I recently left my position as CFO & Director at the investment firm The Family, where the last year and a half has been absurdly busy - too busy alas to post regularly. 

While figuring out my next full-time gig, I figured I might as well get back to writing. I have no idea what the format will look like compared to previous Chasing Paper editions, but there are tons of things I haven’t had the opportunity to talk about, so here it goes. 

As a refresher, Chasing Paper is a newsletter giving my perspectives on finance, strategy, tech & innovation - and how they interact.

If you haven’t, subscribe now:

Subscribe now

Enjoy!

Seven years ago (sic), I wrote what I then considered a thought exercise: The future Da Vinci is a robot, on how AI could gradually invade the most human activity of all: Art.

Little did I imagine that in just a few years time, I’d get to play with generative text-to-image AI applications (DALL-E, Stable Diffusion, Craiyon, Wonder and the likes) with my 3-year-old niece prompting stuff like “pirates wearing dresses and eating marshmallows” or “rabbits playing a football game” (both illustrating this article).

Incidentally, ChatGPT seems to be all the rage nowadays, with people using it for a wide range of unexpected applications, especially by implementing it in their work routine, such as writing, reviewing or commenting code, or adding it to their daily workflow.

Since ChatGPT is trained with human feedback (Reinforcement Learning from Human Feedback), and millions of humans already conversed with it, I thought it would be fun to ask ChatGPT itself some ways it has been used in daily work for other users. 

Here is its mesmerizing reply:

As revealed, plenty of the daily tasks of millions of qualified workers can already be outsourced to a free (for now) AI chatbot. Those advances are astounding, and yet we’re barely scratching the surface of what AI will be able to do.

Copywriting, CRM, marketing, content creation, product or software development, financial analysis, even consulting or teaching could all be radically transformed by implementing trained algorithms into our daily workflows.

While we’re still far from ubiquitous AI in our worklife, for a large number of reasons (high server usage and energy consumption, accessibility barriers, lack of close-ended applications, questionable UI, etc.), the pace at which we’re advancing is remarkable, as revealed by the anecdote opening this article.

A lot of entrepreneurs are now building applications on top of OpenAI or other AI and machine learning platforms (Layer 1s), which is both very exciting and thought-provoking.

With a new Golden Age of AI advancing, we’re likely to see a swarm of hyped-up, VC-funded companies that will end up being wrong bets (think ChatGPT-based customer chatbots, illness diagnosis engines, homework support tutors, etc.).

But that’s not the main issue: more importantly, we should question the irruption of AI into our lives.

Specifically, we should consider what biases AI brings along when replying to our prompts. Some users have already revealed what blatant biases, discriminations or errors are inherently present within the algorithm, that is without even factoring in the malicious attempts at tricking ChatGPT to learn more biases or false truths (which by the way, even if for information purposes and to evidence biases, only makes it worse).

As impressive as the new AI applications are, it should be remembered that learning algorithms are only as good as the data they’re fed with. Notably because for now, AI assistants will not look for the truth or correct answer when prompted with a query ; rather they will seek to provide the answer that sticks the most to what its database (say, the Web) or the people correcting its learning algorithm have said previously.

Put simply, the algorithm knows that on average, the answer it gives is the expected answer to the query - or it infers it from output data.

Apart from the ethic considerations - which are tremendously important but not my point here - trusting the algorithm too much could have important adverse business implications and risks, such as:

• Decreased differentiation among competitors: imagine having the same exact marketing elements, publications, product developments, ideas than competitors. AI could very easily lead one to blend in rather than stand out. Same goes for every application of generative AI: cover letters, blog post content, song lyrics, whatever usually comes from human inspiration.

• Misinterpreting off-pattern signals: if used in an analytical capacity, AI could lead to misinterpreting signals it has never seen before. Especially, those signals could very well be indicating a crisis or conversely product-market fit. How will managers act when incorrectly advised by their algorithm? Will they transfer some of the risk to the algorithm (much like they already do consulting firms) or will they recognize the algorithm might misinterpret something?

• Reduced radical innovation: by suggesting things it has seen somewhere else, AI does indeed help with creator’s block, but it also suppresses any form of newness and radical innovation. How long before relying on it becomes a liability rather than an asset and locks incumbents into a stream of failing strategies (because already washed out)?

Of course, not all businesses and activities need to be unique and unpredictable. Some are purely execution-driven. And while we’re in the infancy of work-related generative AI, many will make fortunes arbitraging opportunities and catching low-hanging fruits.

My take however is that the tools to radically improve execution output thanks to AI will come and generalize, sooner than we expect. 

When this comes, humans will be required more than ever : we have an incredible capacity at breaking patterns, creating new things, innovating - which will all be the key to business success, and to not falling into AI-induced biases that could set us back rather than push us forward.

In the meantime, if you have cool AI apps, I’m all for it!

Cheers,

Younes

Share

Chasing Paper is a newsletter giving my perspectives on finance, strategy, tech & innovation - and how they interact. It’s aimed at CFOs and other people with entrepreneurial endeavors who are looking for tools and ideas to improve the handling of finances in their companies. 

I’ve hit the pause button on content for a while now. However, this week and the coming ones, I have a few complex topics that I need to gain deeper understanding about, and I’d like to take this newsletter as an excuse/opportunity to explore them. Consider it a thought experiment. The format will likely be shorter, more unstructured than other Chasing Paper editions. 

As you will see, those topics are pivotal to my firm’s current and future positioning - and as such, all your ideas and suggestions are welcome!

Enjoy!

I’ve recently started watching Marvel’s Loki on Disney+. The show is quite entertaining if you’re an MCU enthusiast (if you’re not, my guess is that it would be...difficult to follow). The plot revolves around a thought-provoking hypothesis: although there are multiple universes (multiverses), there is a single, sacred timeline determining them all. However, some errors sometimes slip through the cracks of the sacred timeline. They are called variants.

That, for the show, is when the drama kicks in, because it’s precisely what allows for Loki to understand he can exploit timeline variations to his own advantage. And for my purposes here, it’s when an interesting question arises: if market timing is everything, then surely building a time-variant business model is (i) difficult but (ii) very rewarding. 

What I call a time-variant business model is a business abnormality, a model that should not have happened ; and yet managed to exist, thrive or succeed because it relied on favorable timing to market and exploited timeline variances. And very often those businesses are financial businesses.

I can think of one case, very close to home. My firm The Family is a result of what you could call a ‘startup ecosystem arbitraging scheme’. Back in 2013 when Alice, Oussama and Nicolas first envisioned it, the French startup ecosystem was nowhere near where it is now: it lacked capital, it lacked talent (both to create and join startups), and the level of entrepreneurial knowledge was low. But somewhere, on a different timeline, was a startup ecosystem that was thriving: Silicon Valley. What The Family did was to take advantage of its knowledge, network and connections from an advanced ecosystem as well as its understanding of local specificities and hurdles to create a unique model: a safe space for French entrepreneurs that could begin leveling the playing field by leveraging the best startup ecosystem out there. A lot of people called The Family an accelerator then. It was in fact a time-variant community, living in 2 realities at the same time.

The Rocket Internet model, as designed by the famous Samwer brothers, is another example of time-variance: replicating successful business models established in the US in emergent regions requires living in several realities at the same time.

Building such a model is difficult: not only does the playbook need to be honed and executed perfectly, but the model is also ephemeral by essence. Time-variant opportunities do not last forever. That is the reason why many time-variant business models falter.

And yet, many financial institutions started as time-variant models. One of the reasons why the Rothschild family has so many successful banking branches across the world traces back to when Mayer Rothschild decided to send his 5 sons across Europe to set up shop. The one rule was that they each should write to all the others daily. Having daily news from Frankfurt, London, Paris, Naples & Vienna was an incredible competitive advantage over other purveyors of capital, especially in times of war between European kings. They were able to master the time-variance of their individual universes.

More interestingly perhaps than their mere existence, how do time-variant businesses become forever institutions? 

I’ll kill the suspense: I’m not sure. But I think it’s by embracing their business model’s ephemeral nature, exploiting existing time-variances to extract capital, and then reinvesting that capital in subsequent time-variances, staying a certain number of steps ahead of irrelevance as the normal course of time sets back in.

Where are the exploitable time-variances now? I think the global pandemic has created many, many fragmentations in our societies. Entrepreneurs everywhere, especially in emerging markets, are bursting with ambition but are still faced with lack of capital, lack of talent, lack of proper local infrastructure or education. In the meantime, many of those needed resources have become commoditized, available everywhere online, remotely. Managing to create a business model connecting those elements is a unique opportunity at exploiting a massive time-variance.  

As Drake and Future put it: What a time to be alive!

Until next time,

Younes

If you enjoyed, share this newsletter and make sure to subscribe if you haven’t👇

Subscribe now

Share

Chasing Paper is a newsletter giving my perspectives on finance, strategy, tech & innovation - and how they interact. It’s aimed at CFOs and other people with entrepreneurial endeavors who are looking for tools and ideas to improve the handling of finances in their companies. 

Many of you joined the community after the last issue was sent out so let me bid you a warm welcome! And to the others, welcome back!

I have been caught up by many an (actual) CFO duty lately, which is the reason why no issues have been sent. And with everything that’s happened in the finance world since then, I have plenty of stuff to talk about.

For starters, this week I wanted to explore the crypto space from a CFO’s perspective: namely what can and should be done within this (not so) nascent space.

I first heard about cryptocurrencies in 2011. My dear friend Alex Hamou kept ranting about his investments in Bitcoin (huh?), using obscure words such as ‘protocols’, ‘blockchain’, ‘decentralized’, etc. I didn’t pay much attention, since he had a history of not-so-great alternative investments (fortunately, his acumen is better now).

Fast forward 10 years, cryptocurrencies have become ubiquitous for anyone remotely interested in finance, investment or technology. They have become an asset class in which many investors - retail or institutional - deploy capital, and more and more entrepreneurs are launching projects to take decentralized finance further.

And yet not many companies outside of the crypto space have made incursions into what’s called DeFi (Decentralized Finance). CFOs remain skeptical, or interested but ignorant about it all, although there is now clear demand for experience in Bitcoin on the CFO’s skill list.

So as the manager of your company’s finances, what approach should you take?

Master of Coin

In February 2021, Elon Musk decided that Tesla should buy $1.5bn in Bitcoin and that the company would begin accepting Bitcoin as payment. 

Musk being Musk, this decision naturally caused a surge in the price of Bitcoin: tons of retail investors on the Internet love Elon, and anything he says or does has an impact on market prices. At this point, Musk is basically living proof of the signalling theory (if you’re interested, there are many other aspects to the Elon Musk effect, as I wrote in this 2019 piece).

Then in March 2021, Tesla sold about 10% of its total Bitcoin stake, leading to a drop in the cryptocurrency’s market rate. While certain commentators regarded this as a “pump & dump” move (Tesla buys Bitcoin, Elon brags about it on Twitter, price goes up, Tesla sells, price goes down), the Technoking of Tesla has guaranteed that the ultimate motive was to “prove liquidity of Bitcoin as an alternative to holding cash on the balance sheet”. And as his Master of Coin [that’s CFO in the common tongue] Zachary Kikhorn later confirmed in the quarterly earnings call, the company was “quite pleased with how much liquidity there is in the Bitcoin market”. Oh, and they also made $101M in realized capital gains and $929M in unrealized capital gains in the process.

Now, if you’re Tesla and you have an Elon Musk up your sleeve, you would be missing out on a lot of fun (and profits) if you didn’t put this type of power into use regularly. Imagine being able to boost your earnings by a few % just by sending out a few tweets?!

The question remains, for all you non-Tesla CFOs: is Bitcoin a credible alternative to cash in the bank? Is it diligent & prudent for a CFO to hold such an unstable asset on the company’s balance sheet? 

The answer sounds simple, simplistic even: if you’re a crypto-enthusiast, stack up! 

As very well explained by Byrne Hobart in Can an Unstable Asset be a Store of Value, the bull case for Bitcoin dictates that its price will fluctuate violently over time, until it doesn’t:

“The bet for Bitcoin is that a predictable supply will make it a wealth-preserving asset when the money supply rises, but the only way to bootstrap from "free" to "the same market capitalization as gold" is for the price to go up. And prices don't go up in a straight line. Any given Bitcoin price represents the odds that it will replace gold or the dollar as a major reserve asset. But its odds of doing so are also a function of price and volatility: Bitcoin is only as widely-adopted as gold if its market capitalization is equal to gold's, and it's only as good a store of value if its volatility is the same as gold's. (...)

So the paradox of Bitcoin is that the process for becoming a good store of value requires a stage of being a terrible store of value but a very exciting vehicle for speculation.”

In a way, Tesla has found a way to take advantage of both timeframes: it speculates (while saying it doesn’t) in the short term and builds up a position in an asset for which it (has to) believe the price will increase tremendously over the long term.

Note that this applies to believers in the bull case for Bitcoin. Other, more prudent approaches, could lead to devising strategies to hedge your cryptocurrency risk, in the same way you would hedge a regular currency risk.

But above all, I think the debate should be larger than just Bitcoin (or any other single currency for that matter). Similarly, it’s not just a matter of whether or not your company should hold cryptocurrencies as an asset on its balance sheet. Those are mostly questions if you’re an investor (or a bank). 

To me, the important matter is how to leverage Decentralized Finance to your advantage.

Bona DeFi

Decentralized Finance (or DeFi) can be defined as the sum of all cryptocurrencies,  decentralized applications (Dapps) and protocols built on top of a blockchain (usually the Ethereum blockchain) that aim to replicate and improve legacy financial infrastructures (banks, payment gateways, etc.). 

By their very nature, the services offered by DeFi players are interesting for CFOs: they provide innovative, state-of-the-art financial services that can tremendously improve your financial management. 

I would highly recommend reading the excellent guide prepared by the folks at Multis: Breaking Legacy: a CFO’s guide to DeFi & Crypto. It gives an excellent idea of what kind of improvements you can expect from DeFi in your business operations. However, below are some of the direct use cases that can apply to your business which I found interesting.

• Accepting crypto-payments

First things first, the most basic use case of DeFi for any business is to accept payments in crypto-currencies, e.g. with Circle or Coinbase Commerce. 

Now, this is not as straightforward as one might think: Stripe notoriously ended their support for Bitcoin in 2018, arguing that as Bitcoin transitioned from a means of payment to an asset (I mean, one guy did pay for a pizza with 10,000 BTC - now worth > $500M), it didn’t make sense for their customers to accept payments in Bitcoin.

However, more and more companies (including Tesla, as I mentioned above) are accepting payments in Bitcoin and sometimes other cryptos. It’s easy enough to think of the benefits to doing so: instant global payments, no chargebacks, lower transaction fees, etc.

Still, within your business I think there should be a clear combination of factors justifying this: 

• First, you must have customers who are crypto-sensitive, which means they are either in the higher-end of the market (finance and investment-literate) or in the lower-end (locked out of all legacy payment methods because of where they live or their social status).

• Second, you must know whether you want to clear every transaction to fiat currency as soon as possible (and in this case, you would need hedging) or if you want to build up your balance sheet by stacking cryptos. Basically, know if you are going to “hodl” the currency (thereby making it an asset for you) or if you just consider it a means of payment.
  
  By the way, there’s a funny story about the unexpected virtues of having a bunch of Bitcoin sales lying around: there was an urban legend a few years ago that the rapper 50 Cent  became a Bitcoin millionaire by allowing Bitcoin payments for his 2014 album “Animal Ambition”. The story was a hoax, but if you’re bullish on crypto, then you might very well be looking at significant capital gains triggered by Bitcoin-denominated sales. Have a hustler’s ambition, for God’s sake!

• Run business operations

Because Dapps replicate and improve upon legacy financial infrastructure, making the decision to run your business operations on blockchain-fueled services can yield many positive results. 

Basically, you’d be going from fragmented, outdated, slow, centralized, high-fee infrastructures to interoperable, innovative, instantaneous, decentralized, low-fee ones. 

Again, the guide put together by Multis - Breaking Legacy: a CFO’s guide to DeFi & Crypto - is an excellent one to navigate all the use cases (and associated risks) when doing so.

In a nutshell, you could be running your payroll, earning (high) interest on your savings account, managing your expenses and reconciling earnings, accounting and taxes through layers of connected apps. Bear in mind that blockchain transactions are programmatic, which means - as is very important in Chasing Paper - that you could automate many of the menial CFO tasks by simply updating your infrastructure.

While this all feels very futuristic - including for me - I’m convinced this is the future of FinOps (Financial Operations) for many CFOs, either through DeFi or through application layers built on top of legacy infrastructures (what we call FinTech, and I actually think the two will complement each other).

• Digitize your business

This aspect is at the intersection of finance and what we call ‘product’ in the startup world. 

The capabilities offered by blockchain can be used to propel your core business into the future, and this is particularly true of many financial businesses. 

Loyalty programs, for instance, are an important financial addition to many businesses that can end up being worth more as a standalone business than the initial one. And by their very nature, rewards, points, cashback, etc. are very simple things to tokenize. One could easily imagine a company issuing their loyalty points in the forms of tokens, redeemable for various kinds of rewards. The business could then borrow with the tokens as collateral, make them a marketable security, or many other things. Closer to the ground, early in 2021, Visa has partnered with a DeFi company to start offering bitcoin rewards to new credit card holders.

Finding ways to implement tokens or blockchains in your business, if it makes sense, has a lot of upside. It adds speed and security to your processes while boosting your balance sheet with a new kind of financial security: your own. For instance, if Disney were to issue DisneyCoin in lieu of the late Disney Dollar, they could be a stablecoin (the value of 1 dollar to 1 Disney dollar is always 1) that could help improve many payment processes in their resorts, cruises, etc.; but these could also be NFTs (each Dollar is a limited edition), thereby creating another asset for Disney. 

We are not there yet, but I am sure there will be many unexpected ways to use blockchain or crypto to transform large parts of existing businesses. Already, there are groundbreaking companies like Fairmint (we’ll talk about them more below) leveraging the power of DeFi to turn cost centers into profit centers: look how they made money out of their latest marketing material by turning it into an NFT. Genius! Being able to monetize your marketing spend really upends regular business dynamics, for the better.

• Continuous Fundraising

Finally, another groundbreaking impact that DeFi could have on companies is on the financing side. 

I just mentioned Fairmint, which to me is one of the most exciting companies of the 2020s already: they allow any company in the world to program their own equity and thus be in position to raise funds continuously - not only from investors but from all kinds of stakeholders (e.g. drivers for Uber or merchants for Shopify) - without needing to worry about dilution or paperwork. It’s fantastic!

So in the end, the question for a CFO is much larger than ‘should I invest in BitCoin?’. Sure, being knowledgeable about the space and making a first investment in a given cryptocurrency is where it starts, but there are many ways to leverage DeFi to transform your business, and it’s worth thinking about. 

Until next time,

Younes

In the papers 🗞

Some finance-related content I enjoyed recently:

• I was invited by CFO Connect to have a chat with their community of CFOs. We discussed many things that are pillars of Chasing Paper, including Finance Hacking. They turned the discussion into an episode of their podcast CFO Yeah!, that you can listen to on Spotify or Apple - or read the highlights here, if like me your brain does not do well with podcasts.

• My firm (The Family) and Spendesk joined forces to build The Early-Stage Startup Finance Playbook: everything (well, not quite, but most things) there is to know about putting together your finance department when starting a company, including a checklist of simple steps to take. Read along.

• I receive and read many newsletters on a daily/weekly basis, but one I make sure never to miss is Matt Levine’s Money Stuff. Matt’s writing is crisp and his insights about the financial markets and business at large are invaluable. I really recommend subscribing :)

Payment in Kind 🎁

If you enjoyed, share this newsletter and make sure to subscribe if you haven’t👇

Subscribe now

Share

And now, let’s close with things I’ve found in pop culture that somehow relate to finance. This week: memes.

I’ve argued in the past that memes were a part of pop culture. In fact memes are the symbol of an entirely new culture, born and bred online in the Internet Nation.

What’s really incredible is that memes now have an impact on stock market prices (e.g. ‘Dogecoin is up because it’s funny’). 

Many accounts create and share memeable content about stock markets, corporate finance, cryptocurrencies, etc., and their audience is such that they can affect the general view the public has of a given company. 

Among my favorites, I really recommend Litquidity, Arbitrage.Andy, Ebitdad, PrayingForExits, HoeingForYield, NotYourFathersBroker, and Finance_God. Have fun browsing through them!