None of them lived through any of it the first time. That’s sort of the whole trick - they didn’t inherit our embarrassment about wearing it, so it doesn’t read as dated to them. It just reads as new.

For a while I’ve held a loose thesis: culture and commerce don’t move in circles, and they don’t move in straight lines either. They spiral. Same shapes, recurring on a rough cadence - ten years, twenty, thirty - but never landing in the exact same place twice. Each pass picks up the accumulated understanding, technology, and scar tissue of the last one.

It’s why I decided to name this page Spiral Notions - because it continues to be an underlying thesis of… almost everything for me.

The mechanism isn’t mysterious. It’s mostly demographic. A generation lives through something, decides they’re done with it, and spends a decade building the alternative. The next generation, having never lived through the original, doesn’t inherit the aversion - they inherit the absence. And absence is exactly what makes something feel fresh again.

I’ve been particularly watching this happen in real time in two places right now: travel, and loyalty.

Travel Is Already Showing the Tell

Search interest in “nostalgia travel” is up over 3,000% year-over-year. Cruise lines are marketing “the golden age of ocean travel” - formalwear, slower itineraries, a ship that behaves like an occasion instead of a shuttle. Vintage train journeys, analog photography tours, guided group trips: all the things a certain generation spent fifteen years rejecting as uncool are being rediscovered as relief.

That’s the tell. Nobody spirals back to something they still have. They spiral back to something they gave up. Or maybe the generations before them did.

We gave up the tour guide for the algorithm. We gave up the travel agent for the review site. We gave up the ship’s formal dinner for the budget red-eye. Each trade felt like progress - more control, more optionality, more choice. And it was, for a while.

But infinite choice has a ceiling. When every itinerary is self-built, every recommendation AI-generated, every photo already algorithmically optimized before you’ve even taken it - the thing that starts to feel luxurious isn’t more options. It’s someone else having already made the decision for you.

That’s not nostalgia for the 1950s. It’s nostalgia for not having to choose.

Loyalty: The Same Spiral, One Generation Ahead

If travel is where the spiral is visible, loyalty is where it’s structural.

And I've made this argument before - that points, as a mechanism, are dying, and that exclusivity and access are the next version of loyalty. But it's worth pausing on why that's true, because the erosion isn't the interesting part. The interesting part is that it's happened before, on a schedule, and looking at that history is an incredible opportunity to forecast what’s next.

Status, the original model, was relationship-based. Scarce. An agent who knew your name, an upgrade granted at a gate agent’s discretion, value that lived in a person’s memory rather than a database. It didn’t scale, which was the point - it was supposed to feel earned and rare.

Points broke that open. Loyalty became a currency: transferable, hoardable, engineered around. An entire hobbyist economy then formed around gaming it well. It scaled beautifully - which is also why it’s now breaking.

The evidence is piling up fast. Hyatt just moved its top-category redemption from 45,000 to 75,000 points a night - a 67% repricing, overnight. British Airways killed its tier-point system entirely this year in favor of a revenue-based model. Air Canada just sold 25% of its Aeroplan program to an investor group (still sitting with that one). Hotel points liabilities outstanding are north of $11 billion, and every program has a balance-sheet incentive to shrink what it owes you before you can spend it.

Points are dying the same way status died: the mechanism that made it feel valuable stopped being credible.

So what comes back?

Not status, exactly - nobody’s un-inventing the internet, and nobody’s bringing back the gate agent who remembers your name. But the shape rhymes. What’s emerging looks like paid, immediate access: subscription passes, metal-card tiers you buy rather than earn, “lifestyle rewards” that pay out now instead of promising a flight three years out. It’s status’s original promise - be treated well today, not eventually - delivered through a cash mechanic instead of a relationship mechanic.

Three eras, one underlying complaint being answered three different ways:

Status said: earn it through relationship, wait for recognition. Points said: earn it through repetition, wait for redemption. What’s next says: stop waiting. Pay for it, get it now.

Each era solved for the failure mode of the one before it - status didn’t scale, so points scaled it; points got gamed and devalued, so the next model removes the wait entirely. That’s the spiral, not a circle. We’re not returning to 1985. We’re arriving at something that answers the same question 1985 asked, with everything we’ve learned since.

Why the Cadence Matters

This is the part worth sitting with if you’re allocating around it rather than just observing it: the spiral isn’t random, and it isn’t annual. It moves on a generational clock, which is why the 10-20-30 year cadence shows up so consistently across categories - fashion, design, and now travel and loyalty on roughly the same schedule. (I particularly love comparing this to Japanese Fashion…for another post…)

That has a practical implication. If you can identify what a category gave up under duress - not what people say they want, but what they were forced to trade away when the previous model hit its ceiling - you can make a reasonable bet on what gets rebuilt, and roughly when. Points didn’t die because people stopped wanting rewards. They died because the mechanism stopped being trustworthy. The rebuild will restore trust, not restore points.

So the question isn’t “what’s coming back?” That framing makes it sound like nostalgia, a soft trend to note and move past.

The better question is what was given up under duress that’s now overdue for a rebuild - because that’s not a trend, it’s a gap with a waiting market attached to it.

The challenge for investors isn’t spotting the nostalgia. It’s underwriting the rebuild before the category admits it needs one - while the incumbents are still busy defending the model that’s already failing.

For the last twenty years, travel has operated under a fairly accepted premise: distribution is king.

The suppliers had the inventory. The aggregators had the demand. And over time, the aggregators won.

Companies like Booking Holdings and Expedia Group built extraordinary businesses because they understood something earlier than the suppliers did: owning the customer journey mattered more than owning the product itself.

Hotels had rooms. Airlines had seats. But OTAs had attention.

They mastered search, performance marketing, UX, merchandising, conversion optimization, and cross-shopping behavior. They became the starting point for travel discovery itself. And in exchange, they extracted enormous economics from suppliers.

To the naked eye, it looked like suppliers had slowly allowed their inventory to become intermediated away from them. The relationship migrated upward into marketplaces while the brands themselves became increasingly commoditized beneath them.

But the reality was always more nuanced than that.

Most revenue managers would tell you the same thing: empty inventory is worthless inventory. A hotel room unsold tonight disappears forever tomorrow morning. Airlines, hotels, cruises - they all operate against perishability. So if an intermediary could reliably fill distressed inventory, the economics often still worked, even after commissions.

The industry accepted the trade: margin in exchange for occupancy, control in exchange for demand. And for a long time, that trade made sense.

The Internet Is Fragmenting Again…Or Is It?

One of the dominant narratives right now is that AI will fragment the internet even further.

Consumers will no longer move linearly through Google into supplier websites. Discovery will happen everywhere: inside LLMs, voice agents, creator ecosystems, banking apps, messaging threads, operating systems, and embedded commerce environments. Consumers will simply express intent and expect the internet to assemble itself around them.

And on the surface, that appears highly fragmented.

But I increasingly wonder if we are misunderstanding what fragmentation at the interface layer actually means economically.

Because fragmentation at the surface does not necessarily mean fragmentation underneath.

In fact, AI may ultimately create another era of centralization…and even more importanly, the opportunity for it.

The OTA era centralized demand because consumers needed a place to aggregate information. Search costs were high, comparison was painful, and inventory was fragmented. OTAs solved that problem extraordinarily well. They reduced friction, simplified comparison, and built trust into online travel purchasing.

But AI solves discovery too.

Which raises a much more uncomfortable question for suppliers: if consumers increasingly rely on AI to make decisions on their behalf, why would they suddenly revert back to supplier-first behavior?

Why would consumers open twenty hotel websites again if an AI agent can instantly compare, contextualize, assemble, and transact across all of them simultaneously?

The more powerful the agent becomes, the less likely the consumer is to leave the agent.

And structurally, that begins to resemble the OTA dynamic all over again.

The Next Aggregators May Not Look Like OTAs

This is why the current wave of “AI booking infrastructure” companies feels so important - and why I think the market is still underestimating what is actually happening.

Dozens, maybe hundreds, of startups are positioning themselves as some version of the connective layer between suppliers and LLMs. AI-native booking orchestration, agentic commerce, travel copilots, LLM booking rails, autonomous booking infrastructure. The language changes slightly, but the underlying ambition is similar: own the new transaction layer emerging around AI.

But the deeper question is not whether booking moves into AI. It almost certainly will. (Though the proof is not here…quite yet! LLMs are clearly still figuring out their user experience…)

The real question is who captures aggregation once it does.

Because history suggests aggregation is where the power accrues.

The best AI booking systems improve with scale. More inventory improves recommendations. More transaction history improves prediction. More servicing data improves reliability. More user behavior improves personalization. More cross-platform visibility improves optimization.

These systems become more intelligent the larger they become.

That dynamic feels much closer to marketplace economics than people currently acknowledge.

Which is why I’m increasingly skeptical that hundreds of fragmented AI travel agents ultimately survive independently. Consumers may experiment broadly at first, but internet behavior historically consolidates around whichever platforms produce the best outcomes with the least friction.

That consolidation has happened repeatedly across search, social, rideshare, food delivery, e-commerce, and travel itself. AI may not break that pattern. It may accelerate it.

The Incumbents May Be Stronger Than People Think

Right now, there’s a tendency to assume that AI resets the board entirely - that incumbents become vulnerable and new entrants capture the next wave.

Sometimes that happens. But travel may be more structurally difficult than people realize.

Because travel is not just a discovery problem.

It’s a servicing problem, a trust problem, an exception-management problem, and a global operations problem. Booking the trip is often the easiest part. Handling everything that happens after the booking is where complexity explodes.

Refunds, changes, cancellations, disruptions, re-accommodation, customer service, loyalty integration, fraud, payments, localization, supplier connectivity - these are not lightweight infrastructure layers. They are operational systems built over decades.

And incumbents already possess them.

That’s why I increasingly think many of the newer AI booking infrastructure companies may be underestimating how difficult travel infrastructure actually is. A lot of the innovation today is concentrated around the shopping moment - the inspiration layer, the planning layer, the conversational layer.

But travel economics are built across the full lifecycle of the trip.

The companies that ultimately win the AI era in travel may not necessarily be the companies with the best interface. They may be the companies that best combine consumer trust, supplier relationships, servicing infrastructure, transaction scale, and proprietary behavioral data.

And today, many of those companies already exist.

But Suppliers Have One Advantage This Time

At the same time, suppliers are not entering this cycle from the same position they were in twenty years ago.

Historically, suppliers were technologically disadvantaged. OTAs genuinely built better transactional experiences than the brands themselves.

That gap is now compressing.

AI-native tooling is making it dramatically easier for suppliers to improve personalization, merchandising, and conversion optimization on owned channels. Companies like Variant Now hint at a future where supplier websites become adaptive systems rather than static storefronts - the adaptive web.

The same hotel website may subtly reshape itself depending on who arrives: different imagery, different sequencing, different emotional framing, different room merchandising, different loyalty prompts.

The website stops behaving like a brochure and starts behaving more like a responsive sales engine.

That matters because suppliers may finally regain some of the conversion sophistication that OTAs historically dominated.

But more importantly, suppliers may increasingly connect directly into AI ecosystems themselves.

The next generation of “direct” may not involve the consumer visiting Marriott.com at all. Instead, suppliers may expose inventory directly into AI environments while still controlling pricing, merchandising, loyalty, and customer relationships underneath the interaction.

And that creates a fascinating possibility: the future may simultaneously centralize and decentralize at the same time.

Consumer behavior may centralize around a small number of AI interfaces, while transaction infrastructure underneath those interfaces becomes more distributed back toward suppliers themselves.

So Who Wins?

I increasingly think the industry is asking the wrong question.

The question is not whether AI kills OTAs.

The question is whether AI changes where aggregation happens.

And my suspicion is that aggregation remains enormously powerful, maybe even more so.

The internet has historically rewarded whoever sits closest to consumer intent. First it was portals. Then search engines. Then marketplaces. Then app ecosystems.

AI agents may simply become the next aggregation layer.

Which means the companies that win may not necessarily be the companies with the best AI models. They may be the companies that best combine distribution, trust, supplier connectivity, transaction infrastructure, servicing capability, and proprietary behavioral data into one integrated system.

In other words, the companies best positioned for the AI era may look surprisingly familiar.

The irony is that after years of fragmentation, AI may not dismantle the aggregation economy at all.

It may rebuild it in an even stronger form.

But that raises the harder question: if loyalty is becoming invisible infrastructure, what does the consumer-facing version of it look like? What do people actually feel when the new loyalty works?

To answer that, you have to go back further. Before points. Before tiers.

You have to go back to status.

The Arc Nobody Is Talking About

Loyalty has moved in a clear arc - and most people are only watching one chapter of it.

Status was the first language of loyalty. The black card. The platinum tier. The velvet rope at check-in. These programs weren’t really about rewards; they were about hierarchy. They told you who you were relative to everyone else. Brands manufactured scarcity, and consumers competed for it. It worked because it was legible: you either had it or you didn’t.

Then came points. The democratization of loyalty. Suddenly everyone could earn their way to something. Frequent flyer miles. Hotel nights. Cashback. It felt like progress - more inclusive, more transactional, more rational. And for a while, it was.

But democratization has a ceiling. When everyone has points, points mean nothing.

Airlines can devalue miles overnight. Hotel chains quietly inflate redemption thresholds. Credit card rewards get clawed back or capped. The currency printed itself into irrelevance. Points didn’t replace status - they just delayed the next evolution.

We may be approaching the death of points as the primary loyalty mechanism. And the question nobody has answered yet is: what replaces them?

The Generation That Broke the Model

Gen Z and Alpha grew up inside the points era. They understand the mechanics better than any cohort before them. They stack programs, arbitrage benefits, and treat loyalty like a game - because it is one.

And yet they feel nothing for it.

Unlike Boomers and Gen X, who assigned genuine meaning to status symbols - the upgrade, the lounge access, the recognition at the front desk, those milestones were erroded and commoditized. Younger consumers now treat points as utility, not identity. They are optimizers, not loyalists. They will earn the points and switch anyway.

But here is the thing that gets missed: this generation deeply wants to feel special. They want to feel chosen. They want proximity to things they care about and the sense that they are genuinely inside something, not just a number in a database.

They don’t want a tier. They want a relationship.

The desire for exclusivity never died. The form it takes had to evolve.

Trends Don’t Die. They Spiral.

My purpose in this “spiral notions theme” is that there is a pattern in how consumer culture moves that I keep coming back to: trends don’t disappear, they spiral. They return evolved; in a form shaped by everything that came between.

Status came back after the mass-market era - but as streetwear drops and waitlists, not country clubs.

Vinyl came back after streaming - but as collector culture and limited pressings, not a rejection of digital.

The desire for exclusivity and direct connection is coming back after the points era - but not as platinum tiers and velvet ropes. It is coming back as Access, evolved.

The New Architecture: Access

Access is not a perk. It is the entire value proposition.

Where status said you are better than others, and points said you have earned something, Access says you are close to something that matters.

It is qualitative, not quantitative. You cannot measure it in miles or nights or cashback percentages. It shows up as:

• The ability to message a founder directly.

• A listening session with an artist before the album drops.

• A room that only 300 people in the world have ever been in.

• A community where the person who built the thing actually shows up.

• A drop you didn’t find - you were chosen for.

This is why fan clubs are having a renaissance. Why Discord servers with direct creator access outperform traditional loyalty apps. Why a handwritten note from a brand converts harder than a free upgrade. Why the most coveted experiences on earth right now are not the most expensive ones - they are the most closed ones.

The new exclusivity is not about what you can afford. It is about how close you are.

What This Means for Brands

The implications are structural, not cosmetic.

Legacy loyalty was designed to reward spend and became highly quantifiable. Access-based loyalty has to reward relationship - showing up early, going deep, demonstrating genuine affinity. That requires a fundamentally different operating model.

It means:

Curation over scale. Access programs are, by definition, small. The value comes from the limitation. Brands that try to Access-wash a mass program will fail immediately - consumers can feel the difference between genuine proximity and manufactured scarcity.

Qualitative benefits over quantitative ones. The new currency is not points. It is moments. Experiences that cannot be purchased elsewhere, that exist because of the relationship, not despite the transaction.

Direct connection as the product. The most powerful thing a brand or creator can offer right now is themselves - their time, their attention, their genuine presence. The brands that figure out how to architect that at even modest scale will own a disproportionate share of consumer trust.

Earned, not bought. Access cannot be unlocked by swiping a card 50 times. The most defensible form of it is earned through engagement, passion, and community membership - not spend. This shifts the loyalty dynamic entirely.

Loyalty Is Not a Contract or a Connection. It’s a Feeling of Proximity.

In Part I, I ended with this: Loyalty is no longer a contract. It is a connection.

I want to push that further.

Connection is still too passive. The next generation of loyalty is not about being connected to a brand. It is about feeling close to something that matters to the user - an artist, a community, a world, an idea. It is about the difference between following and belonging.

Status gave people hierarchy. Points gave people utility. Access gives people proximity.

The arc is not linear. It spirals. And we are at the beginning of this next turn.

I fundamentally believe that one of the strongest feeders - especially in the AI era - is personal access to technology.

Not institutions. Individuals.

Individuals are the ones who make new technology real. They experiment before there are defined use cases, integrate tools into their daily workflows, and generate the feedback loops that make something intuitive. Over time, that behavior pulls institutions in.

This pattern has presented itself before.

The internet didn’t begin as an enterprise system. It was messy, decentralized, and shaped by individuals exploring it in real time. Only later did companies reorganize around it.

The same was true with mobile. Consumers adopted smartphones first, and by the time enterprises adapted, expectations had already shifted. Everything needed to be immediate, on-demand, and accessible.

AI is following that same arc.

Before enterprises formed strategies or governance frameworks, individuals were already using it to write, think, and problem-solve. Not perfectly, but consistently enough to build familiarity.

And that familiarity mattered - not just because it drove adoption, but because it translated the technology. It made AI usable.

Institutions didn’t introduce AI to people.
People introduced AI to institutions.

From Tools to Systems of Agents

But institutional adoption is not just scaled individual usage.

At the individual level, AI behaves like a tool. The loop is tight: input, output, improvement. The value is immediate and visible.

Inside an institution, that loop expands into something fundamentally different.

AI becomes a system - or more precisely, a network of agents and sub-agents that must operate together. Different models handle different tasks. Outputs from one system feed into another. Decisions are made across layers, not in isolation. And humans are (at least for now) layered on top!

The challenge is no longer whether a model works. It’s whether everything around it works together.

Because once AI is embedded into workflows, it stops being a tool and starts behaving like infrastructure - something that must be reliable, auditable, and consistent across teams.

What looks like a productivity unlock at the surface becomes an orchestration problem underneath.

And orchestration does not move at the same speed as individual usage.

Two Productivity Curves

AI is often described as a universal productivity unlock. As if it lifts all participants at the same rate.

But in practice, productivity is splitting.

At the individual level, gains are immediate. A task that once took hours now takes minutes. The feedback loop is short, and the improvement is tangible.

At the institutional level, productivity is slower to surface. It shows up in how systems connect, how decisions improve, and how workflows evolve over time.

An individual can become more productive overnight. An institution cannot - because institutional productivity depends on coordination, not just capability.

Expectations Reset Before Systems Catch Up

What makes this moment distinct is that individuals didn’t just adopt AI early - they reset expectations.

Employees now enter organizations already AI-enabled. They’ve experienced speed, leverage, and automation firsthand. They don’t see AI as optional. They see it as obvious.

But inside companies, the reality is different.

Organizations are still securing data, evaluating vendors, defining policies, and building internal infrastructure. They are solving for reliability, not just possibility.

Some of them are also prohibiting certain tools, and only allowing for others.

This creates a gap - between what individuals know is possible and what institutions can consistently deliver.

That gap is where friction builds. It’s also where some of the most important opportunities will emerge.

A Travel Example

Travel makes this dynamic clear.

On the consumer side, the experience is collapsing into a single interface. Planning, optimizing, and booking can increasingly happen in one flow. What once required multiple tools and intermediaries now happens in a single interaction.

On the institutional side, the problem is entirely different.

Airlines, hotels, and platforms are not just improving a touchpoint. They are coordinating systems - revenue management, distribution, personalization, inventory, operations - each increasingly powered by different models, agents, and data flows.

This is not a prompt problem. It is an orchestration problem.

And orchestration requires more than better models. It requires infrastructure that can support multiple layers of intelligence operating simultaneously. It requires systems stable enough to be trusted, even as they become more dynamic. It requires data flowing cleanly across environments, and outputs that can be relied on across teams.

But above all, it requires something that hasn’t fully existed before:

Employees working alongside AI - not as users of a tool, but as participants in a system. Designing how agents interact, setting constraints, supervising outcomes, and adapting to a world where execution is no longer fully human.

That shift is not just technical. It’s organizational.

And in a sector as fragmented as travel, it’s what makes orchestration both the challenge - and the opportunity.

Where Value Actually Accrues

If productivity is splitting, so is value.

The individual layer drives adoption and behavior change. It moves quickly, but it is also more replaceable.

The institutional layer is slower, harder, and more complex - but once it works, it compounds. It becomes embedded in how an organization operates.

One moves faster.
The other lasts longer.

From Intelligence to Coordination

The question is no longer whether AI will drive productivity.

It’s where that productivity becomes durable.

The companies that win won’t just build better interfaces or more powerful models. They will build systems that can coordinate intelligence - across agents, across workflows, and across the organization itself.

Because in this cycle, intelligence is becoming abundant.

Coordination is what will be scarce.

The K-shaped economy is no longer just a macroeconomic concept. It’s becoming an opportunity for more defined product strategy and ICP.

Over the past few months, we’ve heard endlessly about the “K-Shape” - some pulling ahead with more spending power, others becoming more constrained. What I think is happening now is wildly structural, and will thread into other categories. That divergence is starting to surface in how people use technology, not just how much they spend.

I see the future of consumer tech as K-shaped.

At the top of the “K,” the consumer is prioritizing refinement, taste, and outcomes. This is a user who doesn’t want more choice - they want better, more distilled choices. Increasingly, they are willing to pay for curation, for personalization, and for someone (or something) to narrow the world on their behalf.

At the bottom of the “K,” the consumer is prioritizing cost, control, and flexibility. This user is willing to do the work themselves if it means a better, faster - but most importantly, cheaper - outcome. Technology here isn’t about abstraction it’s about empowerment.

Across both sides, consumers are turning to technology as a kind of second brain - something that gives them access to information, compresses time, and improves decision-making. But where the real divergence shows up is in cost.

If you want a human in the loop - judgment, taste, accountability - it’s going to cost more.
If you’re willing to replace that with software, you can dramatically reduce that cost.

What’s interesting is that both of these consumers are being served by better products than what existed before - but in fundamentally different ways.

A clean example: Travel

Travel is one of the clearest places this shows up, given how tightly it sits at the intersection of technology and consumer behavior.

On one side, platforms like Fora Travel are building around the idea that the best experience comes from layering human judgment and curated networks on top of technology. The product is not just access to inventory - it’s access to taste, relationships, and time savings. You’re not just buying a flight or a hotel - you’re buying confidence in the outcome, a relationship, and a human layer of accountability.

But that layer comes at a cost - whether explicit or embedded.

On the other side, companies like Mindtrip lean into fully self-directed planning. The user becomes the travel agent, with AI handling the heavy lifting. It’s faster, cheaper, and puts control directly in the hands of the consumer.

Both are working - but for entirely different users, at different ends of the K.

The middle is getting squeezed

For a long time, consumer tech has tried to sit in the middle - building products that serve everyone, abstract complexity, and maximize scale. I think that strategy is becoming increasingly fragile.

The middle is getting squeezed.

Part of that is economic. As affordability becomes more top of mind, consumers are making more explicit trade-offs - between cost and convenience, between software and service.

The next generation of companies won’t be the ones that unify the market. They’ll be the ones that intentionally pick a side.

Build for the top of the K, and you’re building for higher spend, stronger retention, and differentiated experience - but with more operational complexity and cost. Potentially higher LTV, but structurally lower margins.

Build for the bottom, and you’re building for scale, efficiency, and product-led growth - but with more competition and lower pricing power. Potentially lower LTV, but higher margins.

AI changes the equation

Historically, venture has favored the latter. Scalable, software-driven, mass-market products have dominated. But that assumption may be worth revisiting.

AI is changing the equation.

For the first time, it’s possible to deliver something that looks like service at scale. That blurs the line between the two sides of the K. It raises the question of whether the top of the K - traditionally constrained by human capital - can become significantly more scalable, and whether the cost gap between the two sides begins to compress.

Where this leaves us

So the question isn’t just where the opportunity is. It’s where it endures.

The bottom of the K offers larger TAM and faster growth.
The top offers stronger monetization and deeper defensibility.

But what feels increasingly clear to me is this:

The risk is no longer choosing the wrong side of the K.
The risk is trying to build for the middle.

Consumer behavior isn’t converging. It’s splitting.

And the companies that win will be the ones that decide - early and explicitly - where they belong.

In my time working with large enterprises, helping chauffeur innovation into organizations that are - by design - built to move carefully, it’s clear the path is rarely straightforward. These are real, operating businesses with real P&Ls, real org charts, and real constraints. Not startups. Not labs. Companies that have to make technology decisions that stick. And ones that preserve their market presence.

Over the last two years, most of those conversations have followed a familiar arc.

2024 was curiosity.
2025 became the year of “Okay, but can AI do for me?” - and, crucially, prove the ROI. Pilots turned into deployments. Experiments turned into budget lines. The question shifted from “Is this real?” to “Show me where it pays.” And we found…it paid off.

Now, 2026 feels like something different.

2026 is shaping up to be the year of:
“How do I put people in place, who can in turn put AI in place?”

Not what the tools are. Not which model is best this quarter. But: Who inside the organization (or who can I hire that) knows how to operationalize this? Who can translate strategy into systems, workflows, incentives, governance, and change management? Who can take “we should use AI” and turn it into “this is now how work actually gets done here”?

At the same time, there’s this low-grade anxiety in the background. About public markets. About multiples. About whether we’re early or late. About whether this is a platform shift, a bubble, or both. Everyone is trying to squint into the fog and predict how this AI inflection point unfolds.

And that instinct - to predict - is completely understandable. When the ground is moving, we want a map.

Historically, in moments like this, we default to narrative-building. We argue about winners and losers. We debate second-order effects before first-order ones have even settled. We try to place bets on the future.

But in practice, prediction is usually a distraction from preparation.

You don’t control how interfaces evolve.
You don’t control which platforms end up owning distribution.
You don’t control whether agents replace forms, or whether “websites” as we know them slowly dissolve into APIs and machine-readable endpoints.

What you do control is whether your organization is structurally ready for multiple plausible futures.

Discovery in an AI-Mediated World

One of the clearest places this tension shows up is in how companies are thinking about digital discovery.

More and more teams are asking some version of:
“If AI becomes the front door to the internet, what happens to us?”
“If answers are synthesized instead of clicked, what happens to traffic?”
“Do we still build landing pages?”

What they’re really asking for is a prediction. A clean answer. A stable map for a very unstable territory.

Now we can name the anxiety more explicitly: what happens if systems like OpenAI, or other LLM-driven interfaces, become the primary way people find information? What happens if agents, not humans, are increasingly the ones “visiting” your business?

This is where the SEO conversation usually surfaces as a proxy for something bigger.

The more honest answer, though, is: nobody actually knows how this fully plays out.

What is clearer is the direction of travel. Discovery is becoming more machine-mediated. More answers are being synthesized. More interfaces are being abstracted away from “ten blue links.” More agents and models are sitting between users and your brand.

So the strategic question isn’t “Will SEO die?”
It’s: “Is our digital presence prepared to be consumed by machines as well as humans?”

For most enterprises today, the answer is… not yet. (But likely soon…)

Their content is written for persuasion, not retrieval.
Their data is locked in PDFs, CMS silos, or brittle internal systems.
Their sites assume a human browsing journey, not an agent calling an endpoint.
Their “digital strategy” still quietly assumes a screen, a query box, and a click.

Preparing for an AI-mediated world looks very different from trying to predict which interface wins.

It looks like:

• Structuring your data so models can actually understand and use it

• Thinking in APIs and machine-readable surfaces, not just pages

• Treating content less like copy and more like knowledge infrastructure

• Designing for retrieval and action, not just ranking

• Assuming multiple agents, models, and tools will touch your business—whether you planned for it or not

In other words: don’t try to predict whether the landing page dies. Prepare for a world where your company is increasingly accessed programmatically - by both humans and agents.

Zooming back out, this is the broader shift I see heading into 2026.

2025 was about proving AI works.
2026 is about building organizations that can live with it. Indefinitely.

That’s a people problem before it’s a technology problem.

It’s about hiring and empowering operators who can sit at the intersection of product, data, ops, and change management. It’s about rethinking incentives, risk, and ownership. It’s about accepting that “AI strategy” is not a deck - it’s an operating model.

The companies that win won’t be the ones with the best prediction about how this all shakes out. They’ll be the ones who quietly, methodically:

• Put the right people in place

• Rewired how decisions get made

• Rebuilt how work flows through the company

• And made themselves adaptable to futures they can’t fully name yet

The uncomfortable truth is: most of the important outcomes of this transition are not predictable in advance.

But they are preparable for.

And in moments like this, preparation beats prophecy every time.

Every time technology changes how work gets done, we reach for the same vocabulary. We say something is “over,” “broken,” or “dead,” long before we agree on what’s actually replacing it.

SaaS is in one of those moments.

The phrase “SaaS is dead” shows up everywhere right now - not because software has stopped mattering, but because the way we’ve talked about software for the last decade no longer fits what it’s becoming. The category hasn’t collapsed. The mental model has. (Though, I will give credit to the fact the opposite is currently being reflected in the markets…)

When Seats Became a Stand-In for Value

For years, the industry’s dominant story was simple: sell access to tools. More users meant more value. More seats meant more revenue. Growth could be plotted against headcount, and pricing mapped cleanly onto org charts. It wasn’t philosophically elegant, but it was operationally convenient. The seat became the stand-in for value.

Pricing models, go-to-market motions, and competitive moats quietly formed around that proxy. Expansion followed hiring. Stickiness followed sprawl. And investors learned to read the same signals - net retention, seat growth, ARR compounding - as evidence of product strength and durable advantage.

That shortcut worked - until software stopped being just a tool.

AI Changes Who Does the Work

AI doesn’t primarily make people faster at clicking buttons. It changes who, or what, is doing the work in the first place. When systems can draft, decide, execute, and close the loop, the number of humans logging in becomes a weak signal for the amount of economic value being created. A company can move more volume with fewer people, and the software can be doing more while being “used” less, at least in the traditional sense.

This is where the old pricing logic starts to strain.

A Concrete Example: Travel and Revenue Management

You can already see this in travel. Take revenue management. The old model was software that suggested prices and scenarios, with humans reviewing, approving, and pushing changes live. You paid for the number of users who needed access to the tool.

The new model is increasingly automated. Systems test prices, adjust inventory, respond to demand shifts, and optimize in near real time within defined guardrails. One or two people might supervise the system, but the software is doing the work of what used to be an entire team. In that world, pricing by seat starts to feel disconnected from what’s actually being purchased: continuous optimization and revenue impact. The ROI has changed.

From Systems That Support Work to Systems That Do It

This creates a strange tension. The product is delivering more impact, but the pricing model is still anchored to a world where impact and headcount moved in lockstep.

What’s actually happening is a quiet shift in the role software plays inside organizations. We’ve spent decades building systems that record information and systems that help people operate. Now we’re starting to see systems that act. They don’t just surface options; they choose within constraints. They don’t just queue work; they complete it. The human moves from operator to supervisor, from executor to architect of the rules. When execution leaves the human, the seat quietly disappears with it.

Why the Economics Have to Follow

Once you see software that way, the economics start to reorient.

If the system is doing work, then value shows up in throughput, not logins. In tasks completed, not seats provisioned. In dollars saved or generated, not in how many people have access to the interface. Pricing naturally drifts toward usage, volume, and outcomes - not because subscriptions are obsolete, but because access is no longer the scarce input.

Where the “SaaS Is Dead” Narrative Really Comes From

This is where the “SaaS is dead” narrative gets its emotional energy. A lot of companies were built for a world where software’s job was to support humans, not replace or compress their labor. In that world, selling seats made sense.

In a world where software increasingly behaves like a worker, it starts to feel like pricing forklifts by how many people touch the controls rather than by how much freight they move.

What This Means for the Future

If the unit of value is shifting, then a lot of familiar assumptions go with it.

We should expect pricing models to keep drifting away from access and toward activity - away from seats and toward throughput, outcomes, and impact. Not because subscriptions disappear, but because subscriptions alone stop explaining where value is actually created.

We should also expect company-building to change. The most valuable software won’t just be the cleanest interface or the most configurable workflow. It will be the system that can be trusted to act: to make decisions within constraints, to execute reliably, and to improve with use. The moat won’t be how many users you have, but how much real work flows through your product - and how painful it would be to take that system away.

And we should expect org charts to keep thinning in some places and thickening in others. As software absorbs more execution, humans move up the stack: defining goals, setting policy, handling ambiguity. The center of gravity shifts from operating the machine to designing it.

In that light, “SaaS is dead” reads less like a diagnosis and more like a category error. Software isn’t going away. It’s becoming more economically central, not less. But the way we package, price, and evaluate it will keep changing as its role changes.

The real question isn’t whether SaaS survives this transition.

It’s whether our mental models of software evolve fast enough to keep up with what it’s actually becoming.

That world is gone.

What’s emerging in its place is not the end of loyalty - but a fundamental redefinition of what loyalty means, who it serves, and how it is architected.

Two forces are colliding:

1. A generational shift in how people relate to brands, ownership, and commitment.

2. A technological shift that makes real-time personalization, portability, and embedded experiences the baseline expectation.

Together, they are transforming loyalty from a closed, brand-centric system into an open, intent-driven, networked layer of the travel experience.

The Generational Break That Proves It

Gen Z and Millennials are now the fastest-growing share of travel spend, and their relationship with loyalty looks nothing like that of the generations that built today’s programs.

They are more active with points.
They are enrolled in more programs.
They understand the mechanics better than any cohort before them.

And yet - they are less loyal!

Only ~38% of Gen Z report having a “go-to” brand, versus more than 50% of Boomers. They switch for novelty at multiples of older generations. They value flexibility over exclusivity, access over ownership, optionality over lock-in. They’re using more points, and points programs, but disperced across more brands.

This is not irrational behavior. It is a rational response to abundance.

When discovery is infinite, when comparison is frictionless, when AI surfaces alternatives instantly, loyalty is no longer about committing in advance. It is about being relevant in the moment of intent.

Loyalty in the Age of AI

At the same time, consumer expectations have been reset by daily interaction with intelligent systems.

When people can ask a model anything and receive personalized, contextual, real-time responses, they begin to expect the same from every brand interaction. Static tiers, batch-processed offers, and rules engines designed decades ago simply cannot compete.

Personalization is no longer a premium feature.
It is table stakes.

This creates a structural mismatch:

Legacy loyalty was built for repeat behavior.
Modern consumers operate on adaptive decision-making.

Legacy systems are optimized for accounting.
Modern expectations are optimized for experience.

From Brand-First to Intent-First

Historically, loyalty sat at the top of the funnel. You chose a brand first, then planned within its ecosystem.

Today, the journey runs in reverse.

A traveler begins with intent:
“I need a hotel in Austin.”
“I want a weekend escape.”
“I want a wellness experience.”

Search - now increasingly conversational and AI-mediated - surfaces options. And the habits from this new form of search are contagious. Loyalty becomes one signal among many, not the default entry point.

In an intent-first world, loyalty can no longer rely on captivity. It must earn relevance continuously.

The New Architecture of Loyalty

What emerges from this shift is not weaker loyalty, but a different form of it - one that is:

Networked, not siloed.
Portable currencies, cross-brand utility, multi-partner ecosystems.

Real-time, not batch.
API-first infrastructure, dynamic offers, context-aware engagement.

Embedded, not bolted on.
Integrated into payments, wallets, discovery flows, and experiences - not confined to standalone portals.

Personal, not programmatic.
AI-driven understanding of individual travelers, not cohort-based segmentation.

Younger travelers are already “gaming” systems to create this reality themselves: pooling points, stacking benefits, arbitraging across programs, treating loyalty as a multiplayer environment rather than a monogamous contract.

The infrastructure simply hasn’t caught up - yet.

Loyalty Isn’t Dying. It’s Becoming Infrastructure.

In the same way that cloud quietly replaced on-prem, and mobile quietly replaced desktop, loyalty is shifting from a visible “program” to an invisible layer of the travel stack - embedded, intelligent, adaptive, and always on.

The winners will not be the brands with the most points.

They will be the ones whose systems understand intent, reward behavior in real time, and move with the traveler - not try to contain them.

Loyalty is no longer a contract.
It is a connection.

The most durable impact of robotics on commercial real estate will not come from robots that look like robots. It will come from systems that behave like them.

In the built world, robotics is less about form factor and more about function. It’s about technology that can perceive an environment, interpret signals, make decisions, and execute actions across physical space - with or without a visible body. Software that senses occupancy, adjusts energy usage, routes maintenance, coordinates logistics, or responds dynamically to real-world conditions is already doing robotic work. The “robot” isn’t always a machine moving down a hallway; sometimes it’s an orchestration layer quietly coordinating dozens of physical outcomes at once.

This distinction matters, because it explains why some robotics efforts fail - and why others quietly succeed.

Consider consumer-facing robotics like Roomba. The vision was compelling: a physical device that automates a mundane task in the home. But success depended on near-perfect execution in messy, unstructured environments - irregular layouts, edge cases, pets, cables, furniture, constant human interference - all while selling a one-off piece of hardware with limited margin and little recurring leverage. The robot had to do everything itself. When it failed, there was no system (or human) underneath to absorb the error.

Now contrast that with where robotics is working: autonomous vehicles.

Robotic cars are succeeding not because driving is simple - it isn’t - but because the environment is constrained, rule-based, and increasingly sensorized. Roads are standardized. Lanes are painted. Traffic follows protocols. Vehicles operate within defined geofenced areas, with dense mapping, redundancy, and constant feedback loops. Most importantly, autonomy doesn’t exist in isolation; it’s embedded in a broader system of infrastructure, software, and human fallback.

In other words, robotic vehicles don’t rely on a single device performing flawlessly in chaos. They rely on systems that perceive, decide, and act across a structured physical environment.

That same systems logic is what makes the built world fertile ground for robotics.

Hotels, offices, retail, logistics hubs, and multifamily properties are structured, repeatable, and operationally dense. They generate constant streams of data and require thousands of micro-decisions every day: when to clean, cool, repair, route, restock, secure, or staff. The challenge isn’t performing one task perfectly - it’s coordinating many tasks continuously, under constraint.

That coordination problem is where robotics, broadly defined, actually thrives.

In these environments, robotics succeeds not by replacing humans wholesale, but by automating decision-making and response layers beneath them. Sensors, software, and systems work together to perceive what’s happening, determine what should happen next, and trigger action - sometimes through machines, often through people. Humans remain central to judgment, service, creativity, and exception-handling. Robotics absorbs the repetitive, time-sensitive, and coordination-heavy work that otherwise fragments attention and slows response.

This is why the “robots will take over” narrative misses the point. The real value isn’t domination; it’s augmentation. Fewer reactive tasks. Fewer manual handoffs. Fewer blind spots. More leverage for the people already operating these assets.

Seen this way, robotics in the built world is already here - it just doesn’t announce itself loudly, though it has a long way to go. It shows up as systems that act autonomously across physical environments, turning buildings from static assets into responsive platforms. The transformation won’t be cinematic. It will be infrastructural, incremental, and largely invisible.

But that’s exactly why it will last.

Like other foundational technologies before it, once this layer is fully embedded, it won’t feel like a revolution. It will feel like the way buildings were always meant to work.

At the top of the venture hierarchy sit “Tier 1” firms-names like a16z, Sequoia, Accel-whose resources fundamentally alter the risk profile of their portfolio companies. These firms don’t just write checks; they deploy infrastructure. Recruiting teams, policy arms, PR machines, in-house researchers, go-to-market advisors, follow-on capital, and founder networks create an environment where failure becomes statistically harder—not impossible, but meaningfully less likely.

This is not because their companies are inherently better. It’s because the margin for error is wider.

A startup backed by a Tier 1 firm can miss a quarter, reset a narrative, replace leadership, or pivot markets-and still attract capital. Brand alone buys time. Time buys optionality. Optionality buys survival.

Recently, a16z formalized this reality with the launch of what they call “New Media.” The premise is simple but powerful: media, distribution, and narrative are no longer downstream marketing functions-they are core infrastructure. By building owned channels, trusted voices, and cultural reach, a firm can shape markets before companies even enter them. For portfolio companies, this means instant amplification, credibility by association, and a pre-built audience. In practice, it turns attention into leverage-and leverage into defense.

This isn’t marketing. It’s risk mitigation.

Contrast this with companies backed by emerging managers or smaller funds. The same misstep that earns a “learning moment” in a top-tier portfolio can become existential elsewhere. Not because the business is worse, but because the support system is thinner and the market is less forgiving.

In this sense, venture is not a pure meritocracy-it’s a capital-weighted game of compounding advantages. Distribution, signaling, and access matter as much as product-market fit, especially in the early years. Capital doesn’t just fund growth; it manufactures credibility.

This raises an uncomfortable but important question: are Tier 1 firms better pickers, or are they better insurers?

The answer is likely both. Strong pattern recognition matters. But so does the ability to prevent downside, shape perception, and finance through volatility. When a firm can lead, re-lead, price, defend, and narrative-manage its companies, outcomes skew upward almost by design.

For founders, this reality reframes fundraising. Capital is not neutral. Who you raise from determines not just valuation, but survivability. For LPs, it reframes risk. Top-tier funds may appear to outperform because they are playing a structurally different game-one where losses are actively suppressed.

And for emerging managers, the challenge is clear: unfair advantages exist—but they can be built differently. Sector depth, speed, trust, and conviction can become alternative moats.

Venture rewards the appearance of inevitability. Tier 1 firms don’t just find it—they engineer it.

Nowhere is this more visible than in stablecoins. Adoption today is split between early believers and hardened skeptics. Yet when we interrogate where stablecoins’ potential success can-and, in my opinion, will-come from, the answer is not technical superiority alone, but mass participation. That participation is often sparked by a real behavioral desire (here, decentralization and programmability), but sustained by something far less ideological: inertia. What enters motion, stays in motion. Stablecoins may not be strictly necessary-but neither were many of the financial tools we now treat as indispensable. So where does the phenomenon actually come from?

History gives us a clear pattern.

The stock market began as a narrow tool for merchants and industrialists to pool capital for risky ventures, with ownership concentrated among a small financial elite. Over time, mass participation through pensions, 401(k)s, and index funds transformed it into a core social and economic institution. Millions of everyday investors now supply constant, automated capital, shifting markets away from pure fundamentals and toward flows, scale, and narrative. Liquidity is deeper and access is easier, but herding behavior and momentum matter more than patience. At the end of the day, a stock is simply a digital claim on a company-or on many companies via an index-its value pegged to performance and collective belief, not unlike a stablecoin tied to underlying assets and trust.

Homeownership followed a similar arc. Through government support, long-dated mortgages, tax incentives, and securitization, shelter became a financial instrument owned, levered, and traded-directly or indirectly-by the masses. Constant inflows and policy backstops transformed housing into economic infrastructure. Prices now reflect credit availability, sentiment, and liquidity as much as wages or rents. At scale, housing behaves less like a place to live and more like a collectively supported store of value.

Government bonds may be the clearest example. What began as a way to finance wars and infrastructure is now a mass-participation asset held by nearly everyone through pensions, banks, insurers, and central banks. Sovereign debt is sustained less by yield than by trust, liquidity, and the assumption of continuity. Functionally, it operates as a globally accepted, institutionally backed store of value-the stablecoin of the nation-state.

When we question how normalized financial tools come into existence, stablecoins begin to look less like an anomaly and more like the next iteration of a familiar pattern. They are not winning because they are perfect. They are winning because participation compounds, belief scales, and eventually, utility follows. In markets, legitimacy is rarely declared-it’s absorbed.

The dot-com era wasn’t just about Pets.com. It was about fiber. And in today’s “are we in a bubble?” frenzy, we’re not talking about that enough.

Let’s look back.

In the late 1990s, investors believed data traffic would grow exponentially - and they were right. What they got wrong was timing. Billions of dollars poured into laying fiber-optic cable across the U.S. and globally. Capacity exploded. Demand didn’t. And demand didn’t fail because the technology wasn’t needed - it failed because there was no immediate ROI engine to pull it forward. That distinction matters.

Prices collapsed. Companies went bankrupt. “Dark fiber” became shorthand for excess and hubris.

And yet, that overbuild became the backbone of the modern internet. Cloud computing, streaming, SaaS, mobile, AI - none of it exists without the infrastructure that was deemed “too early” and “too expensive” at the time. The failure wasn’t technological. It was financial coordination.

Fast forward to today. AI compute, data centers, energy, and model training costs are being built at a scale that looks wildly disconnected from near-term revenue. Companies are committing to enormous, long-dated infrastructure obligations while monetization ramps far more slowly. Critics call it a bubble. But that framing assumes short-term payoff is the goal.

It rarely is in true platform shifts.

This is the same mistake made during the fiber era: judging long-horizon infrastructure with short-horizon financial lenses. The mismatch isn’t belief - it’s duration. Capital structures are often designed for faster returns than the technology can realistically deliver.

So, are we in a bubble?

This is the wrong question. And perhaps the better one is whether we are over-indexing on short-term financial signals to judge long-horizon infrastructure builds.

A bubble implies something hollow. This moment is the opposite. Demand is real. Usage is real. The tooling is already changing how work gets done. What’s uncertain isn’t whether this infrastructure will matter - it’s when the economics catch up to the build, and whether our capital structures, timelines, and expectations are designed to withstand that gap.

A better metaphor than a bubble is a construction zone. It’s noisy. It’s capital-intensive. Progress feels slower than promised. Costs are front-loaded, benefits back-loaded. And from the outside, it looks inefficient - until one day it simply becomes the road everyone uses.

The dot-com fiber boom didn’t pay off on schedule. But it paid off permanently.

The risk today isn’t that we’re building nothing. It’s that we’re expecting the payoff to arrive faster than infrastructure ever does.

History suggests that when the build is real, the returns don’t disappear - they just arrive on a longer curve than the capital markets prefer.

So where does this land us?

Bubbles imply emptiness. This moment isn’t hollow - it’s early. The dot-com era’s overbuilt fiber networks looked like excess until they quietly became the foundation of the modern internet. Will some investments or products fail? Of course. But today’s AI compute and energy buildouts follow the same pattern: real infrastructure, real demand, and a payoff curve that arrives later than fund cycles are designed to accommodate.

The challenge for investors isn’t avoiding the build - it’s aligning capital structures, exit timing, and return expectations with technologies that compound over decades, not quarters.

• Outer is aspiring founders, new angel investors, junior VC analysts, and those just entering the ecosystem. Access to capital, deal flow, and top networks is limited. Quality is a 1/50 bet. 2% of investments drive fund returns.

• Middle is experienced founders, repeat entrepreneurs, established VCs, and those with a few successful exits or investments. These have more access, reputation, and deal flow but are the top echelon. Quality is a 1/20 bet. 5% of investments drive fund returns.

• Inner is elite, well-connected, and successful founders and investors- unicorn founders, top-tier VC partners, and those with significant influence, access to the best deals, and the ability to set industry trends. Quality is a 1/5 bet. 20% of investments will produce fund returns.

Being in the inner circle means you produce better returns and it takes less work to get there.

Because venture is 1. so network driven and 2. a numbers game, you must be in the inner circle to consistently, repetitively win. Consistent, repetitive wins leads to trust. Trust builds firms (LP commitments). Firms build longevity and institutionality.