Think about an agentic system where you feed the system with an initial hypothesis. The first agent is a 'Builder Agent'. That agent builds the first MVP.

It then feeds it to a set of 'n' Customer Twin Agents and they each provide simulated customer feedback on that specific MVP. The feedback includes actual clicks and usage simulation as well as qualitative elements such as quotes about the MVP.

The Builder Agents then iterate the MVP into V2. and launch it again for the Twin Agents to give feedback. In the model we define a probability threshold where we define when it's time to stop the AI-discovery experiment with high conviction for the real validation with real customers.

The Customer Twin Agents also are not fully theoretical - if a company has already customers, we build a layer of customer data from surveys, interviews and data from real customer to train our agents. That means you constantly evolve your Twin Agents to represents what real customers want and behave.

I will present a simulation of how that works in a session I will give as part of the "Introduction to AI-Driven Lean Startups" where I will cover the 3A's framework.

Register here

Over 300 people from some of the top startups and corporations have now signed up to my upcoming session "Introduction to AI-Driven Lean Startup" on Maven.

Product Managers and Innovation specialists from Lego, Amazon, Carta, Swiss Airlines and Salesforce are coming to learn and see how lean startup is evolving with AI.

In the session, I will share my thinking around the 3A's framework on the strategies to apply lean startup in your ventures and products. This session is relevant if you are a startup founder, Product Manager, L&D professional and even if you are in academia researching AI and entrepreneurship.

AI is reshaping how startups discover opportunities, validate ideas, and build products. Founders who combine Lean Startup principles with AI can move faster, test smarter, and make better decisions with fewer resources. This session introduces practical frameworks and tools to help builders turn AI into a competitive advantage.

Here is what we will cover:
- How Lean Startup is changing with AI
- Introduction to the 3A's AI Lean Startup Framework
- Demo of Lean Startup AI tools
- Questions and future courses

Join me on Tue, Aug 25, 2026 5:00 PM GMT+1 (45 minutes)

Only 10 days to go!

The session is FREE, takes 30 minutes with 15 minutes Q&A at the end. If you cannot attend I'll send you a video of the session.

I intentionally aimed to have it in a time where founders and builders from both Europe and the US can join!

Register Today

Two weeks ago I went with my wife to watch a movie at our local Everyman. It’s not something we do often, raising three little boys who are endlessly energetic and demanding doesn’t leave much room for it. But Aviv’s mom was visiting from Canada, we got the kids down early in bed, and for once we made it out. The film ended up giving me a lot more to think about than I expected.

“The Invite” starring Seth Rogen, Olivia Wilde, Penélope Cruz and Edward Norton is a sharp, funny film set in San Francisco. Joe (Rogen) and Angela (Wilde) are a married couple with a young child. Angela invites their upstairs neighbours, Hawk and Pina, over for dinner, and the evening takes a turn none of them saw coming. Joe dreamed of being a musician and life took him somewhere else instead - he became a music professor and suffers from severed back pains. Angela loves architecture and design an redesigned their home (which was Joe’s parents house where he grew up in). What starts as a normal dinner unravels into a night of raw emotion, old frustrations, and an invitation of a very different kind - you’ll have to see it for yourself.

But there’s one line, from Pina to Joe near the end, that stuck with me. “You can choose not to be miserable,” she tells him. It’s simple, and it’s true. Our lives are shaped by how we choose to tell our own story (to ourselves first and then to others). We all have our wins and our failures, our good days and the days we just want to lie down and do nothing. But every single day, we get to choose not to be miserable - to learn from what’s hard, deal with it, and try again.

That’s my decision number one.

Decision number two comes right after it. Every morning, I choose to build a better future than today. I like sitting with those two words together, building and future, because building means action. It means moving one step further toward what we’re capable of and focusing on the future leaves us with optimism and progress.

For me, “today I’m going to build a better future” shows up in a few different forms: building products and ventures, investing in and working alongside founders building the next generation of AI companies, teaching future entrepreneurial leaders at London Business School, and researching new frameworks to help leaders, scientists and CEOs build faster and more effectively with AI.

The most common misunderstanding about building is that it’s synonymous with coding, or engineering. It isn’t. You can build a community, or a business, without writing a single line of code, without even writing a single prompt. Building is making something out of nothing. That’s where I find the most energy: the act of creating something that didn’t exist in the world before.

What building the future means for you is your call. But if you commit to building, and you commit to doing it in service of a better future, you’ll find you use your time far more wisely.

So there it is: two simple decisions for you to make every single morning.

You guessed it: he struggled to close his A round for months and months.

The first problem was that VCs didn’t know how to position it within their investment thesis. If they wanted to lead, should they write a $6m check for a $10m round, or a $3m check for a $5m round? The range was so wide that VCs couldn’t see a clear path to fit it into their investment strategy. It completely confused some of the investors, or made them immediately rejection the deal.

The second problem was that it signaled to most VCs that the founder didn’t actually know how much money he wanted to raise (which is partly true). But the real issue wasn’t that he lacked a number - it was that he hadn’t built conviction in his own venture price. Too many hesitations, too many doubts, too many comparisons to other startups meant he wasn’t convinced his company was worth a specific amount. So he gave a range instead - The wider the range the lower the founder’s conviction.

That’s a weak negotiating strategy as a founder. A VC who hears “$5m” as the floor will anchor to it, and it becomes very hard to move them toward the top of your range, especially if you don’t give a horizon for the price to go up but leave it as a range.

, the founder of Uber and today founder and CEO of Atoms, a physical AI startup, shared a tactic he used to reshape the psychology of anchoring in venture fundraising at Uber negotiating with some of the top leading VCs in the Bay Area. He talked about it on a podcast with ’s and .

Travis calls it the “Uncapped Anchor.” He tells the story of one Uber’s investment rounds where he opened negotiations by telling investors: “This is going to be one of the hottest deals in Silicon Valley this year. We don’t know where the price is going to go, but it’s at least [blank].” He didn’t want to end up in the position where he names the high price, the investor names a low one, and they settle in the middle. Instead, he always anchored with “at least X” with plenty of room to go up in the price. Then he’d read the room and the investors motivation: their vibes and the body language of potential investors. He’d say it’s at least 2x the last round. Then 3x. And keep building conviction from there.

You can see how Travis built conviction in his round price incrementally, using the Uncapped Anchor. But even once you’ve built conviction, it can be shaken by decisions of your investors during negotiations. Travis describes pushing the price up and reaching a soft commitment of $375m with Andreessen Horowitz - the highest offer in his process. Then he got an email from Marc Andreessen. Over dinner, Marc told him the new pre-money price they could offer was only $210m.

If you’ve ever negotiated a venture round and had a soft-committed price cut before the term sheet, you know how much that can hurt (a lot!). Travis went back to the rest of his investors and rebuilt conviction from the new anchor of $210m. Eventually, ended up at the top of the list at the time with Menlo Ventures.

Milestone based price conviction (Bottom-up)

When I ask founders why they picked $1m or $5m as their round price, most don’t have a clear, detailed answer. Shockingly, 80% just pick a number that sounds reasonable for their stage. When I push them to do the work, they build a financial and operational projection and then use that number as the pre-money price.

“But what if things go wrong? (And they will.)” I ask. Founder use an easy fix - they add on a 20–25% buffer. That’s the wrong approach in my view. You need to deeply think want to achieve with the capital you raise, by when, and then to use that as a proof to raise your next round and attract the best talent.

My advice to founder is build your price bottom-up, using venture milestones as building blocks. Set your strategic milestones first for example “Launch the first product,” “Reach $50m ARR,” and so on. Then, place them on a timeline, each with its own cost and a running total of the cost to reach the overall milestones. For your buffer, add 3–4 additional milestones beyond your base case, with their own costs. When you add those buffer milestones don’t be shy, be bold. Now you have a bottom-up price built from what you need to hit specific milestones, plus room for expanded ambition through the add-on milestones. You have the timeline, the reasoning, and the price.

Market value conviction (Top-Down)

Setting your round price works the same way as pricing your product. Most founders when picking a product price to charge make the mistake of calculating what it costs to build the product, then pricing to cover that cost. What I push founders to do instead is price around the value the product creates for customers: “How much are customers willing to pay for you to solve their problem?” What’s the real value you deliver? If it’s high, and you can build it at low cost, your margins will be attractive and sufficient.

The same logic applies to your round price. Ask yourself: “How much would an investor pay for my venture?” Not how much you need. Not how much it cost you to build. How much do you think it’s actually worth? think about how much you’d pay to acquire your own startup. This is where Travis’s Uncapped Anchor becomes useful again: it’s a live test of whether investors agree with the price you believe your venture is worth.

This whole process, building your price bottom-up through milestones, then stress-testing it top-down through an uncapped negotiation, should make you uncomfortable. It should create tension, both internal and external with your own beliefs and with your investors’ perspectives. If you don’t push past your comfort zone on price, build your own conviction, and build your investors' conviction, until you reach a number worth paying you will come with a weak negotiation strategy of a wide range that will scare investors away.

Between 2020 and 2022, everyone talked about sustainability and climate tech. That was exactly when I started building in climate tech ventures myself. I joined bp as Entrepreneur-in-Residence to build a new climate and energy venturing unit, and to build new climate ventures from the ground up (0 to 1).

It was the hottest thing in tech.

CEOs were stepping into Chief Sustainability Officer roles, and there wasn’t a single meeting where an executive didn’t mention how ambitious their company’s sustainability goals were.

New climate venture funds were popping up like mushrooms after rain.

Fast forward to 2026: nobody talks about climate anymore, bp Ventures is winding down, and corporate CEOs’ mental capacity is fully occupied by artificial intelligence. That’s the focus now, and that’s where the ambition is headed in almost every company - and there’s a good reason for that.

Since 2023, mentions of the word “climate” on earnings calls have dropped by two-thirds, while S&P 500 companies mention “AI” over 400% more than they did in 2023.

The main reason isn’t that corporates stopped caring about climate, it’s that in 2020–2022, “climate” was a selling strategy. It excited shareholders and pushed public companies’ share prices up; it gave investors confidence and conviction. “AI” is doing exactly the same thing in 2026. Companies that mention AI on earnings calls see their stock prices rise, and they signal relevance, an innovative mindset, and competitive advantage.

Taking a step back and thinking about my long focus on AI and climate, the best companies I know understand the core difference between the two: AI is a tool and probably the most exciting tool humans have created in our lifetime to solve big existential problems. Climate change is one of those big problems humans need to solve.

I believe we’ll keep seeing public companies shift their narratives toward whatever moves the share price. They’ll mention AI a lot more, and maybe sustainability less. But what excites me are the companies that show how they’re using AI to solve climate challenges for the long run, how they deepen their scientific angle and research on with that AI strategy, and how they accelerate that journey.

Climate change is one of the biggest and most important problems our generation of builders and entrepreneurs should solve and AI is one of the most important and transformational tools we need to build in order to solve it.

Disclaimer: the image is an AI-generated illustration comparing the annual average mentions of the terms “AI” and “Climate” (and related terms like ESG) in earnings calls from 2016–2026.

Register Today

Excited to invite founders, investors and innovators to a free introduction to AI-Driven Lean Startup on Maven.

AI is reshaping how startups discover opportunities, validate ideas, and build products. Founders who combine Lean Startup principles with AI can move faster, test smarter, and make better decisions with fewer resources. This session introduces practical frameworks and tools to help builders turn AI into a competitive advantage.

Here is what we will cover:
- How Lean Startup is changing with AI
- Introduction to the 3A's AI Lean Startup Framework
- Demo of Lean Startup AI tools
- Questions and future courses

Join me on Tue, Aug 25, 2026 5:00 PM GMT+1 (45 minutes)

If you cannot join, register to get the recording

Register Today

Some of the most impactful and successful leadership and management mechanisms and principles for startups are inspired by leadership methods applied and developed in large and traditional military organizations.

We like to think of defense departments and militaries as very big and inefficient organizations, but when applied in ventures they are very well aligned with the pace, efficiency and agency startups need.

Chief of Staff

The function of Chief of Staff recently is extremely common from the very early stages of startup companies. The role was orginitated in the military where trusted officers and generals appointed their Chief of Staff to coordinate operations, manage information, and ensure strategic decisions were executed effectively and promptly. This function was adopted later in governments and large corporations. Founders are shaping this role to drive cross-functional execution, managing special projects, aligning teams, and enabling them to focus on vision, customers, and growth.

Squads Governance

The squad organizational entity was originated in the military. It is the smallest tactical unit in the armed forces. It is designed to operate autonomously under a squad commander while coordinating with larger formations and stakeholders such as platoons, companies and battalions. It emphasizes clear objectives, decentralized decision-making, and accountability. Spotify adopted this concept and applied it in the startup scene. They designed organized small, autonomous, cross-functional teams around specific products or customer outcomes. Many startups adopted this approach to increase speed, ownership, and innovation, enabling those small teams to make decisions independently while staying aligned with the company's overall strategy and vision.

Post-Mortem Meetings

Post-mortems originated in the military as structured review processes conducted after special missions or critical projects to investigate chain of evens, reasons for failure, and how future missions could be improved with maximum efficiency and minimum casualties. Google popularized the modern engineering post-mortem process using structured reviews of outages and failures to improve systems, processes, and organizational learning. Post-mortem processes are adopted in startup companies as a way to create a culture of transparency and improvements.

We see other concepts like Deployment Strategist or Deployment Engineers that are taken from military operations.

What are other military principles and mechanisms you apply as founders?

Listen to an AI Generated debate podcast about the topic

It's the third founder I have this tough conversation with in the last 4 months where they all had the same dilemma: to continue building their non-AI startup or pivot to build a new AI venture.

All of the co-founders team are top-performers, outlier founders. They graduated from elite research institutions, worked in the most desirable tech companies and incredibly intelligent and motivated.

They have worked for at least 2 years on their non-AI venture, developing their ideas, validating, trying to find successful business model and raise funding. The main reason they reached out to me for a chat in the last months was that they got a clear message from investors that there is no way they will invest.

My advice to all 3 founders whom I know for some time mentoring them was clear: "It's the time to build an AI venture."

It's a matter of timing in my perspective. As an asset class, venture capital is focusing on finding alpha and investing in the future top performing companies. Today, it's like saying they are looking for top AI companies. I have seen it over and over with my own eyes, how VCs just skip and reject amazing founders, building in large markets just because it's not 'AI'.

It also doesn't have to be a software startup that you are building. You can build a defence venture or a physical AI business (or in plain english robotics company) and you will get the attention you want and need from VCs. VCs have almost 80% of their attention now on AI in 2026. Even if you have all the most important success metrics proven, commercial TRL level, long term paid agreements, amazing growth in an non-AI venture, you might not find the spot in most of the VC portfolios today - that is just the reality I see.

Artificial intelligence firms accounted for 61% of global venture capital (VC) investment in 2025. That is USD 258.7 billion of investment out of a total USD 427.1 billion, more than doubling AI’s share since 2022 (30%) according to new OECD analysis of VC data.

2 out of the 3 founders I mentored who took my advice, who did not manage to raise for their non-AI business for more than a year, raised capital from tier-1 VCs in less than 2 months. I don't think they worked on a bad idea, or targeted a small market it's just that if you want to build a venture backed startup you need to address timing and VC attention.

As much as I hate to see their important research and ventures they worked so hard on moved to the side, with that pivot of actions and mindset they will manage to grow and learn as founders, build an AI solution that will help millions and generate alpha for investors. There might be another time for their other venture where the attention of VCs will go in that direction and maybe by then they will be in a totally different position to make those happen too.

This is the next generation of my new media: 🤙 Illai Gescheit: I write about AI, venture capital, startups, science and Innovation.

and ’s new report about "How tech workers are feeling in 2026" gives a really interesting perspective on what the pace of change in AI but also how companies are managed and governed accordingly.

One of the insights that stood out for me was the increase of burnout of tech employees while their optimism is going down. In 2025, burnout was 44.7% and went up to more than half of the participants in the survey with 55.7% feeling burnt-out. Optimism dropped from 54.8% in 2025 to 48.7% this year. As someone who is working with founders and VCs, you have no choice but to be optimistic or as I consider myself a "Responsible Optimist", but with tech layoff surging as Microsoft laying off 4800 people and shrinking their XBOX division, Meta laying off 8000 people and moving away from its commitment to AI it's difficult for people especially in large tech companies to stay optimistic.

Most people associate burn out with no work-life balance and working too much or with extreme pressure. I think the burn-out shift here stems from uncertainty. The uncertainty is not only about "will I have a job tomorrow?" but mainly is about "what is my job?", "what is the mission and vision of the company with AI?" and "how will I catch up with everything that is going on?". I saw a podcast a while ago with and where he said he was feeling anxious about catching up with the changes daily, so imagine an average tech employee at Cisco, and how they feel with the pace of change - this is a mental burn-out because of the massive acceleration of change.

Years in the industry taught me that the only thing that could solve uncertainty , anxiety and burn-out is a CEO, a leader and a manager who listens and is able to clearly communicate and decide on a vision and direction for the company. This correlates well with the data of how burn-out drops with manager effectiveness. 65% of the responders with higher job enjoyment and dramatically lower burnout. 25.5% of tech employees rate their manager as highly effective. You can argue that an effective manager is able to make good delegation of tasks. I would argue those effective managers are leaders who are able to communicate priorities and direction clearly and frequently.

It keeps the human element as the most important to keep the ship in the right direction and preventing it from sinking. The only remedy for AI burnout are superb human leaders.

SHARE THIS WITH A FOUNDER

This is for you founders - I added today this new feature: All founders can now claim their free Business Model Canvas.

1. Add your website URL.
2. Write your email.
3. Get an AI analysed Business Model Canvas.
4. Share the canvas or download the PDF.

You can later sign up to Between if you wish to edit, collaborate and manage your startup experiments.

Visit https://between.ventures/claim

The SpaceX recent IPO made a lot of news and many people quite wealthy. On June 12, 2026 SpaceX which was founded by Elon Musk went public. The SpaceX public trade was the largest IPO in history and the company was valued US$1.77 trillion making Musk the first trillionaire. The first day of trading went extremely well and the company was traded $2.1 trillion that day, positioning SpaceX as the sixth most valuable U.S.-listed firm. The IPO raised $86 billion that will allow the company to pursue its long-term vision and ambitions. A few days later we see some corrections in the public market valuation which is typical for any IPO.

The SpaceX Optimists and the Skeptics

There are two groups of people around me who take different perspectives on the SpaceX IPO story. The first group of people goes along with the herd and invested in the public round immediately without waiting to see how the market reacts to the stock - you can call them the SpaceX stock first-movers. The lucky ones did it either with the initial stock price of ~$160 per share or when it's ~$153 per SPCX stock after correction. I know a few investors who joined the party when the price was at its peak of over $200 who are less happy about it now but I can assure you they knew the risk. Holding public stocks are a long-term game but this one is quite a unique public investment that feels a bit like a venture deal in the public market playground.

The second group are the skeptics who could not see how they could invest in a venture like SpaceX from a profitability and timelines perspective - the risk and uncertainty are too big. Many saw the valuation as too high for such a deal. Investing in the SpaceX stock feels a bit like investing in the Si-Fi story, as it was written by Isaac Asimov who wrote 'The Last Question', a story published in 1956. Would you invest in Science Fiction as if it is a fact or a proven model?

There were orders of over $100B in shares even from individual investors for the SpaceX IPO while the company planned to raise only $75 billion. It seems like people and institutional investors are investing in a futuristic pitch and reality based on the vision of Elon Musk - in fact they are investing in Musk himself while he is the CEO of Tesla, Neurolink and the Boring Company bringing the question how will he have capacity to lead all those ventures at the same time with public money managed by him.

@carolinehydetv#spacex #ipo got $100B ordered just from individual investors…but most is being allocated to institutions. That’s a lot of demand left unfulfilled? How you bracing for Friday? Lmk ✌️❤️

Tiktok failed to load.

Enable 3rd party cookies or use another browser

What fascinates me is the lack of understanding of the SpaceX business model and the reasoning behind its M&A strategy by existing investors. I surveyed contacts from both of the groups I mentioned above if they are clear about the SpaceX business model. None of them could articulate it well - it was all fluff focusing just on Musk’s ability to build amazing companies and proven track record.

Do you know what is SpaceX's business model? is it space tourism? is it the construction of spacecrafts? is it the data collection? and what is the biggest revenue generating element of the business? What is that company path to profitability? and what are the timelines?

Applying Business Model Canvas on SpaceX

Business Model Canvas is a tool created as part of the Lean startup movement by to help companies on their search for a business model that scales. It has ingredients that any business person, founder and entrepreneur is acquainted with such as key partnerships and revenue streams. Designing all those pieces together on a single canvas, enables us to identify gaps, define hypothesis and iterate the different elements in the business model and understand the dependencies of each part - it allows us to search for a proven and repeatable business model.

It is a bit funny to use a "startup tool" like the Business Model Canvas for a company like SpaceX since it is not designed primarily for companies who have proven business models, but for startup companies in search for their business model as likes to articulate. SpaceX feels like a unique situation in that sense because it feels like it nailed and proved its business model, but this analysis shows us that this might be a fallacy covered by an immense ability for raising capital successfully. It allows us to zoom-out and think strategically of SpaceX ambitions, plans and strategy to form an opinion and world view on the company’s trajectory.

To apply the Business Model Canvas for the SpaceX case, I used the AI-Driven 3A’s framework I have been developing and the Between Ventures product I built. This framework allows anyone to work with AI to accelerate lean startup methods and discovery.

There are three methods to work with AI to accelerate the lean startup method as part of the 3A’s framework. I used Automatic Discovery for the analysis of the SpaceX Business Model Canvas. I used the Between Venture tool to delegate the business model canvas creation to AI. Between Ventures is a product I built that enables you to automate lean startup in two ways. The first method is writing a prompt in English of the business you are building, and the AI will generate the business model canvas for you which you could later edit and iterate man. The second method is by Reverse Engineering a company website as context for the AI to create a business model canvas.

I simply copied the URL of the SpaceX website and let Between Ventures generate the BMC analysis for me.

The output of this analysis is a web based Business model canvas, which I later downloaded the PDF of, and then used another AI tool to make it nicer with the SpaceX design brand inspiration for this article and social post.

The second method in the 3A’s framework, includes building the business model canvas with AI as a thought partner. This is the Augmented approach where you can work with your AI tool of choice such as Gemini or Claude to build your business model canvas manually or directly into a tool like Between Ventures. An example of such AI augmentation includes the user asking the AI tool “if I add to SpaceX NASA as a partner how will that impact my value proposition?”. It’s about step by step construction of the BMC with back and forth discussion with your AI.

The third method is a full Agentic Discovery lean startup approach, where there are sets of agents developing hypotheses, testing and developing again and again until you reach product market fit. Here is an example of such approach where I eventually aim to see Self-Improving agentic Iteration model.

What can we learn from the SpaceX Business Model Canvas

Space Travel Frequency is one of the biggest elements of SpaceX's value proposition analysis in the BMC. When we talk about sending rockets to space we need to differentiate between orbital launches with astronauts on board and ones that help add another satellite to space for connectivity coverage. We have seen a surge or orbital launches since 2002, the year SpaceX was founded. As of 2025, the US has 193 orbital launch attempts, followed by China with 93 attempts.

BryceTech reported that during 2025, there were 325 orbital launches and 4,544 spacecraft deployed in space. This number represents approximately a 25% increase in launches and a 54% increase in spacecraft relative to 2024. As the US remains the dominant force of orbital launching, it stated that SpaceX launched 85% of all satellites in 2025.

74% of SpaceX’s missions were attributed to Starlink with smaller satellite launches. However, we can also think of the vision of sending manned spacecraft more frequently as a long-term value proposition for SpaceX with its multi-planetary value propositio. If we were used to sending a crew to space every 5-7 years, in 2025 there were 8 crewed orbital launches done. Half of them were done by SpaceX (with collaboration with NASA), 2 by China and 2 by Russia. With the vision of multi-planetary life we should expect crewed launches to rise potentially doubling in the next 2 years (8 annual crewed launches).

Everything in the SpaceX business model screams long-term. Looking at the customer segments and key partnerships of the SpaceX BMC, almost every entity the company engages has a long term timeline with procurement, order times etc. The government contracts, the time to form partnerships with partners like NASA and other space agencies. This is relevant in particular for hardware companies who try to optimize short terms contracts - long term is the game.

The SpaceX Prospectus mentions that “We generate revenue in our Connectivity segment primarily through subscription fees from consumer subscribers. We drive consumer revenue through monthly subscription fees based on geographic market and download speed, recognizing revenue ratably over the service period, plus typically a one-time sale of a kit. In addition, we generate revenue from enterprises through contracts structured as a combination of subscriptions, data consumption, and capacity, or on a percentage-of-completion basis, depending on each customer’s particular needs. We generate government revenue via long term contracts for Starshield, a secure satellite network designed specifically for government customers and national security applications. We also earn Starlink Mobile revenue through revenue-sharing arrangements with MNO partners, based on connectivity services included in their plans.”

In some sense, the Starlink and X’s subscription models could create an illusion of a software company generating revenue quickly and growing exponentially like a social media venture, but don’t forget the really long commitments the company as a whole has from servers, satellites and real-estate. Those require the company to optimize for long term contracts with governments and other entities that will balance those expenditures that are of the nature of a rocket manufacturing comapny.

Is SpaceX a hardware company? an AI company? this is one of the key questions that come up when you zoom out and look at this company as a whole. I think the AI BMC analysis missed a core aspect about the company. Between Automatic Discovery completely structured the business model company as a rocket manufacturing company. However, M&A activity with the acquisition and merge of X AI, X and Cursor alongside the Starlink service makes it a hybrid of both.

The SpaceX business is composed of 3 divisions.

The Launch Division is the core business of the space company building spacecraft and launching into space. SpaceX managed to reduce the cost of launching payloads by up to 95% and the company is accounting for approximately 80% of global mass sent into orbit.

The Connectivity Division (Starlink) is the revenue generation arm of the company. Musk is using the increase in frequency of orbital launch of the Launch Division to have more then 10,000 satellites in orbit, with the ambition to scale it to 100,000 satellites in orbit. This part of the business is extremely profitable with 30% margins. This shows about the ambitions to be not only dominant in space orbital launches but also with earth-space connectivity. It also teaches us about the scale of investment needed to get such coverage.

The AI Division (X, X AI and Cursor) are the latest additions to SpaceX with somewhat artificially manufactured acquisition of social media and AI companies by a space company. The recent acquisition of Cursor is joining that M&A strategy with the $60b acquisition that was announced a week ago. This seems like an attempt to compete with what we expect to be the most transformative technology of our time AI, and trying to get into the game with others like OpenAI (maybe for ego sake), Anthropic and Google by Musk. Another angle to see this is the massive compute demand generated by AI companies and the potential operations of data centers in space powered by SpaceX.

Finally, one of the biggest benefits of using the Business Model Canvas is to identify gaps in the business model. If I work backwards from the Between Ventures AI analysed SpaceX BMC, there are endless questions we could ask about how this all comes togethers as a profitable business with a proven and repeatable business model. Starlink alone makes sense, The AI Division might make sense alone as well.

However, the element of access to space (apart of the access to connectivity from space), multi-planetary living strategy is not backed by a clear business and operating models. This alongside losses of $41 billion and immense burn rate and negative free cash flow of about $20 billion raise many question about how a company with so many unknowns regarding it business models could become sustainable and profitable in the near future. It also raises questions around its M&A strategy and the motivations behind merging X, X AI and the acquisition of Cursor.

More questions that come up with the business model analysis are around Musk’s control of the company that deals with governments, militaries and other large entities with high level of regulations, while the company now publicly traded gives super-voting shares that empower Musk 85% voting control, board and CEO controls. In addition, he has legal protections provided by the state of Texas that limit shareholder lawsuits. Can such governance support a complex business as we see in the business model canvas?

I’m Illai. I write about startups, venture capital, Artificial Intelligence and entrepreneurial career paths. You can find more: Follow me on Linkedin| Get updates about workshops and courses | ZIZO Product Course | Between App

Paid Subscribers will get exclusive access to more articles and insights, get an invite to my community and will get discounts for workshops and advisory. Subscribe Today.

From seventh to ninth grade, I had perfect scores in mathematics.

Math was my favourite subject. My two teachers made every lesson feel like a puzzle waiting to be solved. I sat in the front row, did every homework assignment with genuine excitement, and couldn’t wait to see what we’d learn next.

One memory has stayed with me for years.

During one exam, there was a single question I couldn’t solve. It wasn’t difficult, but I couldn’t get it right.

My teacher, who also happened to be the School Principal, told me that I can continue to work on the exam although time was up. He said:

“You have as much time as you need. My office is yours until you solve it.”

Ninety minutes later, I found the solution.

Looking back, he wasn’t teaching me mathematics. He was teaching me perseverance and that he believed in me and my abilities.

Then came tenth grade.

In the 10th grade, I was assigned a new math teacher.

She was exceptionally strict, but what affected me wasn’t her discipline it was her lack of belief in her students. Every lesson felt heavy. I went from looking forward to math class to dreading it. Sometimes I genuinely felt sick before walking into the classroom.

For the first time in my life, I started failing mathematics.

Unfortunately, this wasn’t just another school year.

It was the year that determined which mathematics track each student would follow throughout high school.

In my school there were three levels:

• Level 5 — the highest level, required for advanced physics.

• Level 4 — typically for students pursuing biology or chemistry.

• Level 3 — the basic mathematics track.

I already knew where I wanted to be.

My dream was to study physics at Hemda, a science-focused high school, and eventually become an engineer. To do that, I needed to qualify for Level 5.

So I found a private tutor, who was a friend of the family and knew me well - he also knew that teacher.

He reminded me why I had fallen in love with mathematics in the first place. He made it challenging, exciting, and fun again.

Just before the final placement exam, I approached my teacher.

“I know my average is only 84,” I told her, “but I really want to study Level 5 mathematics.”

She looked at me and said something I’ll never forget.

“You’ll never succeed in mathematics, not even at Level 3.”

For a moment, it felt like my dream had been taken away.

But instead of breaking me, those words became fuel.

I studied harder than I ever had before.

When the placement results arrived, I had done just enough to qualify for Level 5.

I got into the physics program at Hemda.

A year later, I had a different mathematics teacher.

Once again, math became what I always believed it could be: useful, fascinating, and deeply enjoyable.

By the time I graduated high school, I had earned a perfect score in mathematics.

On graduation day, I walked over to my former teacher.

“Remember when you told me I wouldn’t make it?” I asked.

“Well... I did.”

Years later, I remember exactly how those teachers made me feel.

One teacher believed in me before I believed in myself.

Another almost convinced me that I couldn’t succeed.

The difference wasn’t their knowledge of mathematics.

It was their belief.

A single teacher can change the direction of a student’s life not only through what they teach, but through what they choose to believe about the person sitting in front of them.

👋 Hey there, I’m Illai. I share with readers questions, insights and learnings about building startups, venture capital, Artificial Intelligence and entrepreneurial career paths. You can find more: Follow me on Linkedin| Get updates about workshops and courses | ZIZO Product Course | Between App

Paid Subscribers will get exclusive access to more articles and insights, get an invite to my community and will get discounts for workshops and advisory. Subscribe Today.

I talk with founders a lot about sleep.

In fact, I find that sleeping habits is one of the most underrated and undressed topics for founder success.

As a founder, I had very limited sleeping hours from different reasons (every day there was something else), and I feel that every single founder I meet deals with sleep deprivation in one way or another.

I call it "The Founder Sleep Deprivation Paradox". The way people talk about sleep in the entrepreneurship context is on the two extreme of the spectrum. On the one hand there are the VCs and founders who claim you don't need sleep and you have to sleep in the office even if your back is broken.

On the other hand there are people who find sleep very important and say they must get 8 hours of sleep, and are obsessed with that concept.

The reality for me as a founder is messy - The Messy Sleep!

There are three main reasons why I couldn't get enough sleep:

1) Too excited to build - this happened a lot especially with a new launch or when we pitched. I was so excited and eager to progress that the adrenaline was working extra hours. Ideas just pop in my head constantly and they take over sleep - I had it two days ago where I found myself working until 04:00.

2) What if everything falls apart - I worried a lot as a founder. Why does the app crashes so many times? will we get that funding? will we get that approval for partnership? will that customer convert? and today we have so many more strategic and existential things to worry about with AI such as is the impact I'm having is going to do good? will I need to do layoffs now that I can apply AI to the company?

3) Kids, family and everything else - as you grow older, and have kids, other commitments there are plenty of things to be worry about, your kid wakes up at 03:00 coughing, you had to take a red eye flight to get to your son's concert.

Why do I say it's a paradox? the most exciting things in my life such as my kids and family, the problem I worked on like climate and healthcare they all come with great responsibility - so the excitement and drive is mixed with responsibility and sometimes worry.

I told you it's messy.

Could an Oura ring or an Apple watch to track your sleep work? it could help but it can't solve it.

It's a paradox that you have to embrace yourself and understand it exists.

My big learning from spending years in this paradox is that I am able to function well in that messy reality. When I get a good night sleep, I say thank you, and when I get no sleep I leverage it and find the good at it - I found out that some of the most creative ideas I had came to me with very little sleep.

How many hours of sleep you get on average? and how do you react to it?

👋 Hey there, I’m Illai. I share with readers questions, insights and learnings about building startups, venture capital, Artificial Intelligence and entrepreneurial career paths. You can find more: Follow me on Linkedin| Get updates about workshops and courses | ZIZO Product Course | Between App

Paid Subscribers will get exclusive access to more articles and insights, get an invite to my community and will get discounts for workshops and advisory. Subscribe Today.

There are 4 states AI founders and entrepreneurs can find themselves when it comes to building and funding AI Ventures.

There is nothing more clear than position those 4 states on a 2x2 chart.

So if you are building AI, try to identify in which quadrant you are in and write in the comments. You might want to shift from one quadrant to another, or you might learn that you don't want to be in a certain quadrant and pivot to another one.

Here are the 4 states:

1) You are building a in the Foundational/Scientific layer space and needs massive amounts of cash so you raise VC or PE money and lots of it. You are probably building the next AI lab training new models and need that capital for compute power and energy costs. You might work on a big scientific problem in medicine, climate, education or even in defence and you need the best research talent to train models for accuracy and precision.

2) You are building in the application layer space and raise lots of capital - You are probably building for a specific sector like in legal tech, or industrial solutions. You are using existing models like Gemini, GPT or Sonnet or even world models, but you don't need the capital to train your AI model. You need the capital to move fast, and be the first mover in your industry allowing network effects and scalability.

3) You are building in the application space but you bootstrap and don't plan to raise - You might use tools like Codex, Claude, Lovable and others to build your application. It might be a small project you are building to solve a local problem, or you have a conscious decision to keep all equity and move slower. At a certain point where you get a signal for PMF you might want to go up and raise capital where you will move to quadrant 2.

4) You are building a in the Foundational/Scientific layer space but you do not need hundreds of millions in capital - You are probably doing research in academia or even independently, but you are far from building a commercial proposition. You are trying to solve a creative problem and need capital to run your research but scaling it might not be a priority, because you do not have the technical or scientific validity to scale it. You might move to the first quadrant at some point of your research.

Which quadrant are you in and where do you want to be?

👋 Hey there, I’m Illai. I share with readers questions, insights and learnings about building startups, venture capital, Artificial Intelligence and entrepreneurial career paths. You can find more: Follow me on Linkedin| Get updates about workshops and courses | ZIZO Product Course | Between App

Paid Subscribers will get exclusive access to more articles and insights, get an invite to my community and will get discounts for workshops and advisory. Subscribe Today.

What does good looks like in venture capital?

The story of Cerebras and Eclipse can teach us a lot about the Magic of Loyal VC Funding.

Most of the cases I see in startup funding includes a venture fund that enters at seed and maybe follow to series A, but then growth investors come in and take control and dilute other shareholders significantly. Some VCs don't even think about reserves and follow on funding.

The best VCs build long term-partnerships not of one or two rounds but they commit for the right.

Both 's Eclipse and Cerebras took a bet on AI growth before we all thought it's cool. They took a bet that AI would be a huge market to build for, and also that the way to approach it would be from first principles with a new chip architecture that will challenge Moore's law.

Eclipse invested in the the very beginning at series A alongside Benchmark and Foundation through series F - these are some of the hardest moments of building a hardware startup.

Then we see that Cerebras CEO Andrew Feldman was able to form loyal partnership with the growth fund Alpha Wave Global which entered the cap table in Series E just before 2020 and followed on through growth alongside Benchmark in a series H funding of $1.1B at an $8.1B valuation.

This story teaches us a lot about what good looks like in venture funding and mainly these two insights I want founders to take:

1) Find a VC that aligns on the same bets you take on the market, technology and people.

2) Look for long term VC partnerships and optimize for that. You should build personal relationship and make funds a part of a long ride from see to IPO.

👋 Hey there, I’m Illai. I share with readers questions, insights and learnings about building startups, venture capital, Artificial Intelligence and entrepreneurial career paths. You can find more: Follow me on Linkedin| Get updates about workshops and courses | ZIZO Product Course | Between App

Paid Subscribers will get exclusive access to more articles and insights, get an invite to my community and will get discounts for workshops and advisory. Subscribe Today.

I spent almost 7 years of my early career building and designing patent applications and IP portfolio in medTech and health tech. A big part of our startup strategy included an IP strategy.

When I went to build software ventures, I had a conversation with a friend who is a CEO of a large scaleup in the Bay Area, and he immediately told me that spending any money on patents in software is a waste of time and money and that I need just to run fast and execute.

It’s interesting to reflect on that tension between Intellectual Property and software development fro the lens of Machine Learning and Artificial Intelligence. As we have more companies emerge as research labs, IP is an integral part of companies’ strategy including blocking patents and future development in adjacent to research papers and R&D efforts.

In the latest Stanford Human-Centered Artificial Intelligence 2026 AI Index Report we see how AI patents are growing bunt not everywhere:

- The number of granted AI patents has grown exponentially, from 3,866 in 2010 to 131,121 in 2024. Between 2023 and 2024 the number of granted AI patents rose by 8.2%.

- This growth is led by what seems like a strategic move by China which has the majority of 74.2% of the AI granted patents of the global total. The United States is at 12.1% of the granted AI patents, followed by Europe at 3%) and India with only 0.4%.

It looks like we witness two opposite macro-strategies around AI: the western world is diminishing its activity around filing AI patents, and moving faster on the commercial development of AI applications while China is accelerating filing and granting AI patent applications.

The other angle is to see the pace of citations of patents which are granted which show how quickly knowledge diffuses within a discipline. It seems that the pace of conversion between patent grant and citation is faster in US IP granted which could be impacted by translation and language barriers.

I’m trying to reflect on the timeline which a patent gives one protection which is 20 years from the moment you file a patent application. In areas of high regulatory measures like in healthcare, 20 years could be eaten by 10-15 years of trials and then you are left with 5 years of commercial protection while in the pace of development of AI patent application could give protection of at least 15 years.

Enroll to my AI Lean Startup Class

👋 Hey there, I’m Illai. I share with readers questions, insights and learnings about building startups, venture capital, Artificial Intelligence and entrepreneurial career paths. You can find more: Follow me on Linkedin| Get updates about workshops and courses | ZIZO Product Course | Between App

Paid Subscribers will get exclusive access to more articles and insights, get an invite to my community and will get discounts for workshops and advisory. Subscribe Today.

AI is changing how VCs operate and select the best venture deals and it also changes how we teach students about venture capital and entrepreneurship.

the Co-Founder and Managing Partner at BNVT Capital joined us for the London Business School Pathways to Startup Success class. Together with his Co-Founder Rory Mounsey-Heysham, they recently announced the closing of their $150m fund, that invests in 'Benevolent Disruptors' with a unique investment thesis that based on the largest private datasets in venture capital history - 40 years, 500 funds, 14,000 companies.

The students had the opportunity to put on the 'Investor Lenses' and take a role as Associates at BNVT Capital for 3 hours having Chris in the room and AI assistants.

The session was based on a new format developed by Gary Dushnitsky - a case-less case study! this format blends canonical and contemporary themes in entrepreneurship.

We asked important questions such as:
- How do VCs source and screen for entrepreneurial opportunities?
- What makes a rigorous investment thesis and how it could make deal-selection efficient and successful?
- What is the importance of data quality and quantity in the selection process?
- Where should we use AI tools, and when we use our human judgement and conviction?

The simulation in class included several parts:

1) VC success - We introduced venture capital success frameworks.

2) Introduction to Chris and BNVT Capital - Chris shared about his journey as a venture capitalist and what led him to build BNVT Capital.

3) AI-Driven Simulation - based on real data sourced of 50+ companies and market overview, the students had the task of selecting top deals to present to the BNVT Investment Committee.

4) Presenting Selections - based on their deal-selection process, students pitched their deals, shared how they used AI and their prompts.

5) Comparing with reality - the class heard from Chris how his team structured their deal-selection, and how the companies they considered performed.

It was fascinating to see how the class applied AI tools and LLMs, and how that impacted investment criteria, deal-selection processes and eventually the output and deal recommendations.

We also learned how to use AI properly with frameworks such as AI Fluency, prompt engineering best-practices and advanced methods to apply discernment and diligence of AI tools as a VC.

If you want to learn more about the case-less case format, the BNVT Capital simulation and case and about teaching entrepreneurship with AI we recently published the case and the Teaching Note here:

Download the BNVT Capital LBS Case

The pace of AI is evolving fast, and it unlocks new ways in which VCs and founders operate, invest and build ventures and so is the way we teach with AI.

I will share more about teaching entrepreneurship and venture capital with AI in the future.

👋 Hey there, I’m Illai. I share with readers questions, insights and learnings about building startups, venture capital, Artificial Intelligence and entrepreneurial career paths. You can find more: Follow me on Linkedin| Get updates about workshops and courses | ZIZO Product Course | Between App

Paid Subscribers will get exclusive access to more articles and insights, get an invite to my community and will get discounts for workshops and advisory. Subscribe Today.

What can we learn from the 20 biggest LPs who invested in the best performing VCs?

8 of the 20 largest LPs are pension funds. 7 from the US and 1 from Canada. Overall Pension funds with total of 657 VC investments. Next are 5 Funds of Funds 4 US based and 1 in Israel did 557 VC investments. The rest are foundations, endowments and even a corporate (MetLife).

Here are my main 3 takeaways:

No European or British LPs on the list

18 out of the 20 biggest LPs are based in the US, 1 is Canadian and the other is Israeli. Not a single one is European! When we talk about the European VC ecosystem today I definitely see bolder investors and larger deals, there is a huge appetite. However, we also see consolidation of the mega-funds which are US based like at Thrive, a16z and General Catalyst. We do not think about the consolidation of LP money to only US investor which eventually funnel their capital to US funds. Europe must create and build more Limited Partners that invest in successful VCs. That might be the core of the European risk capital change we need in order to make the UK and EU venture market compete.

Israeli Startup Nation has startups, VCs and LPs

Vintage Investment Partners founded Alan Feld is an amazing example of how Israel as the Startup Nation and one of the thriving startup ecosystem was able not only to generate some of the greatest startup deals like Wiz with its $30b acquisition by Google, but also a healthy investment vehicle in funds. Vintage, was able to give access not just to some of the best deals but also to the best funds and when we look it from a system view, the best way to build a thriving and growing ecosystem is to provide access for investors across startup deals, funds, GPs and LPs which all play the Power Law game.

European and British University Endowments should invest in entrepreneurship

When I look at the other categories which invest in the venture class like foundations and endowments the appetite for risk is clear. I found a list of at least 44 endowment funds in top European universities. If we think of an economic growth perspective, European universities are the home of some of the greatest research and technology talents today in the AI, bio and science space. If we think about building and prioritizing more VC money to support entrepreneurial talent coming from academia in Europe the return would be immense. The idea of getting access to those fund managers investing in graduates of building ventures in Europe is a circular view of capital, talent and ideas that could bring an economic change in the AI era.

What are your takeaways?

Enroll to AI Lean Startup Course

👋 Hey there, I’m Illai. I share with readers questions, insights and learnings about building startups, venture capital, Artificial Intelligence and entrepreneurial career paths. You can find more: Follow me on Linkedin| Get updates about workshops and courses | ZIZO Product Course | Between App

Paid Subscribers will get exclusive access to more articles and insights, get an invite to my community and will get discounts for workshops and advisory. Subscribe Today.

Many people don't use printers as they used to, and I try not to use it from efficiency and sustainability reasons.

But for me, as I do this weekend, my printer helps me do focus and deep thinking work.

When I want to do a deep dive, or deeply think about a subject, I define a weekend with less screens and emails, and I print the document I am going to focus on.

My main mechanism is to print the document and then write notes with a pencil.

Sometimes I would read the same document twice, trying to come up with different perspectives and insights - it's an iterative process.

The biggest learning I have doing my deep thinking process is not to let go until I understand every bit fully in the document. In many cases we might skip things and leave it for later.

Today, I might go to AI and try to clarify a point which is unclear. With deep thinking, I will try to analyze and understand deeply everything in the document. We have an urge as humans to give ourselves a break, in that case and move on - in this case I force myself to understand.

Professor Naomi Baron conducted a survey that included 300 university students across the US, Japan, Slovakia and Germany. The students were given a choice to consume content from printed media, smartphones, laptops, electronic books and desktop computers - 92% of the students that the paper version helped them concentrate the most.

AI cognitive work understand the relationships between words (or tokens) but us humans we are visual creatures. We see objects, and the way that we consume text, is as shapes (I clearly saw this when I studies Chinese).

I believe that the creative process of reading and scribbling notes on a paper, is like art to me, and I remember by the physical action of drawing but also with the combination of the way I drew and wrote the notes. I use many forms like small flow charts on the side, formula and more.

Enroll to my new course

👋 Hey there, I’m Illai. I share with readers questions, insights and learnings about building startups, venture capital, Artificial Intelligence and entrepreneurial career paths. You can find more: Follow me on Linkedin| Get updates about workshops and courses | ZIZO Product Course | Between App

Paid Subscribers will get exclusive access to more articles and insights, get an invite to my community and will get discounts for workshops and advisory. Subscribe Today.

Refer a friend

Let me be clear:If you’re building a startup today and you’re not deeply thinking about how to use AI across your company, that needs to change!

Over the past 3 months, I’ve spoken with multiple founders building exciting ventures yet many haven’t fully integrated AI into how they build and operate.

Even if you’re building in deep tech, agtech, climate, or any “non-AI” category AI should and could be embedded across your workflows:

- Validating your business model
- Experimenting with product features
- Running growth and content
- Managing customer interactions
- Preparing for fundraising

Investors will ask you about AI and not just what you’re building but how you think, operate, and execute with it. If you don’t have a clear answer, it signals something deeper and translated as a red flag.

To get you started thinking about AI in the startup building context, here are 8 practical ways founders can apply AI today:

1) Test new product ideas and features with AI coding.

2) Accelerate customer feedback into learning.

3) Make customer service better using agents.

4) Reduce team size by delegating tasks to AI.

5) Use AI to generate content and convert to leads.

6) Summarize users conversations with AI.

7) Build pipelines of potential investors with AI.

8) Keep up to date with competitors by building agents.

This is not optional anymore in the reality of building startups, and not acting today will prevent you from progressing as fast as you could.

👋 Hey there, I’m Illai. I share with readers questions, insights and learnings about building startups, venture capital, Artificial Intelligence and entrepreneurial career paths. You can find more: Follow me on Linkedin| Get updates about workshops and courses | ZIZO Product Course | Between App

Paid Subscribers will get exclusive access to more articles and insights, get an invite to my community and will get discounts for workshops and advisory. Subscribe Today.