Novig raised $75 million in a Series B round led by Pantera Capital in February 2026, with participation from Multicoin Capital, Makers Fund, Edge Equity, and existing investors Forerunner Ventures, NFX, and Perceptive Ventures. The raise brings total funding past $105 million. The company is valued at $500 million.

Plain English: Novig r…

Read more

The Mental Model: First Principles Thinking

Most enterprise AI compliance tools are built on an assumption that compliance is a checklist. You build the thing, then you validate it, then you file evidence, then you move on. The checklist gets heavier with every regulatory update, every audit, every new jurisdiction.

First Principles Thinking says: strip the assumption away. What is compliance actually trying to accomplish? Continuous, auditable evidence that systems operated within defined rules. That's it. If you accept that as the goal, the logical architecture is to build compliance into the execution layer, not bolt it onto the review layer.

That's what Iridius built. And it's why Accenture came in both as an investor and a strategic partner, combining its delivery scale with Iridius's compliance AI infrastructure to help pharma and life sciences clients move beyond isolated pilots. Accenture doesn't make those kinds of bets on seed-stage companies for marketing reasons.

At /mkt, we've been building a version of this same insight into financial services compliance from day one. The FINRA supervisory procedures, the Form NMA workflow, the tZERO integration documentation — none of it was built as an afterthought. It's structural. You can't retrofit a compliance layer into a regulated product and expect it to hold under scrutiny. The regulator will find the seam.

The Contrarian Take

Iridius starts in life sciences, where GxP standards are well-codified and the regulatory frameworks are explicit enough to be machine-readable. That's smart. Start where the rules are clearest.

The harder question is financial services. Banking and capital markets compliance involves more discretionary judgment, more jurisdictional variation, and more contested definitions than pharma regulation does. The same architectural approach applies, but the "machine-readable regulatory logic" is significantly harder to build when the rules themselves are interpretive.

Which makes it a bigger prize for whoever gets there. The compliance-as-execution-layer model is correct. The vertical expansion roadmap will be the real test of whether Iridius is a pharma tool or a platform.

SECTION 1: Company Overview

Adfin is a London-based fintech built to solve a deceptively simple problem: getting small businesses paid on time. But the approach is anything but simple.

The company pairs proprietary payment infrastructure with what it calls “agentic AI workflows”—autonomous agents that decide the best course of action to recover, reconcile, and accelerate invoice payments. Translation: an AI system that acts on behalf of your finance team, deciding which payment method to present, when to retry failed payments, how to sequence reminders, and when to escalate without human judgment bottlenecks.

Read more

The Mental Model: Crossing the Chasm

Geoffrey Moore's framework is one of the most useful in my book. Early adopters embrace something new because they understand the underlying mechanics and can tolerate ambiguity. The mainstream market needs something that works simply, with familiar rules and clear expectations.

Robinhood's venture fund is built for early adopters. The 150,000 sign-ups are people who saw "private market access" and clicked. But the chasm between "people who signed up" and "people who understand what they own, why liquidity is years away, and how this differs from a stock" is enormous.

Crossing that chasm in private markets requires more than a product launch. It requires investor education, clear disclosure, and a structure that actually delivers on the access promise without creating a new category of confused retail exposure.

At /mkt, we've spent 14 months building toward the same access problem from a different angle. Reg A+ qualified securities. A FINRA-regulated ATS. KYC and accredited verification baked into the onboarding flow. The structure is designed to hold up when the mainstream market arrives, not just the early adopters.

The Contrarian Take

Robinhood is right that retail deserves earlier access to private market returns. But 150,000 sign-ups doesn't validate the structure. It validates the demand. Those are different things.

The real test comes in three years when the fund's early positions are still locked up and those 150,000 retail investors want to know when they can get their money back. That's when product-market fit in private markets gets stress-tested.

Demand and readiness are not the same thing.

The Mental Model: Winner-Take-All Markets

Winner-Take-All is one of the models in my book, and this is its most extreme real-world manifestation I've ever seen. The dynamic requires two conditions: network effects that compound with scale, and economies of scale that drive down unit costs as volume grows. Frontier AI labs have both. More compute means better models. Better models attract more users. More users generate more data. More data trains better models. The loop is closed and the distance between the top and everyone else widens every quarter.

The result isn't just that OpenAI or Anthropic win. It's that institutional LP capital, which used to flow across hundreds of sectors and thousands of companies, is now being routed to a handful of platforms that resemble infrastructure utilities more than startups.

The Contrarian Take

Every founder not building in AI is reading these numbers and assuming their raise just got harder. That's the first-order reaction. The second-order read is more interesting.

The WEF report identifies the locked-up capital problem explicitly: companies staying private longer means the exit value that should recycle back to LPs isn't flowing. That capital can't fund the next generation of startups. When the recycling mechanism breaks, the next capital cycle doesn't start.

That means the founders building in less-covered verticals right now are raising in the trough before the rebound. When frontier AI companies eventually IPO and LP capital starts flowing again, the companies that built real infrastructure during the concentration period will be the ones with the regulatory moats, the customer bases, and the distribution that the next wave of capital chases.

At /mkt, we've spent 14 months building regulated infrastructure in a vertical that hasn't seen a single dollar of the AI mega-round wave. That's not a disadvantage. It's positioning.

The winners of the next VC cycle won't be the ones who chased the current headline. They'll be the ones who built the thing the headline missed.

1. True Anomaly Just Raised $650M to Build Space Weapons for the Golden Dome

Series D: $650M
Valuation: $2.2B
Total raised: $1B
Founded: 2022

Colorado-based True Anomaly closed a $650 million Series D this week — co-led by Eclipse and Riot Ventures, with Paradigm, Atreides, and VanEck joining — pushing its total capital raised past $1 billion since its 2022 founding. The timing wasn't accidental: just four days before the announcement, the U.S. Space Force named True Anomaly one of 12 prime contractors on the Golden Dome space-based interceptor program, part of up to $3.2 billion in collective contracts. The company plans to double its headcount from 250 to 500 by year-end, expand manufacturing from 140,000 to 2 million square feet over four years, and fly a dozen missions in the next 18 months — including LEO, GEO, and cislunar operations with its Jackal autonomous orbital vehicle.

HOT TAKE

Four years old, $1 billion raised, zero commercial revenue — True Anomaly's entire business is the bet that space is a warfighting domain before it's a commercial one, and right now the U.S. government agrees.

MENTAL MODEL · THE CLEAN-SHEET ADVANTAGE

Every established defense prime — Lockheed, Raytheon, Northrop — carries decades of legacy requirements, cost-plus contracting habits, and internal politics that slow them down in a new domain. True Anomaly has none of that. They've built every system from scratch for space superiority, which means their architecture isn't constrained by what worked in 1995. The mental model here is the Clean-Sheet Advantage: when a new warfighting domain is being invented in real time, the company with no legacy to protect can outmaneuver incumbents not because it's smarter, but because it's not dragging an anchor. The question for investors and strategists isn't whether the Golden Dome program gets funded — it's whether True Anomaly's clean-sheet speed compounds faster than the primes' institutional weight.

2. Aidoc Raises $150M From Goldman — and Quietly Eyes an IPO

Series E: $150M
Total raised: $520M
Hospitals: ~2,000
Annual cases: 60M+

Aidoc closed a $150 million Series E on April 29 led by Goldman Sachs Growth Equity, with General Catalyst, SoftBank Vision Fund 2, and Nvidia's NVentures joining the round — bringing total funding to over $520 million. The company's CARE foundation model holds 31 FDA clearances, runs on nearly 2,000 hospital systems worldwide, and analyzes more than 60 million patient cases per year; diagnostic errors and delays cost roughly 400,000 American lives annually, the exact problem Aidoc is building against. CEO Elad Walach told Axios the company is eyeing an IPO within three to five years, and Goldman's growth equity team — which has deployed over $13 billion since 2003 — rarely writes a check without a public markets thesis in mind.

HOT TAKE

Goldman leading a healthcare AI round isn't a market signal — it's a market call. When the bank that prices IPOs starts backing the company, it's not early anymore.

3. Legora Just Got Nvidia Money — Making Its $600M Round the One to Watch in Legal Tech

Series D total: $600M
Valuation: $5.6B
Customers: 800+ firms
Markets: 50+

Swedish legal AI platform Legora extended its Series D to $600 million total this week after Nvidia's NVentures backed the company at a $5.6 billion valuation — Atlassian, Adams Street Partners, and Insight also joined the extension. The original $550 million Series D closed in March at $5.55 billion, tripling Legora's valuation in five months from its October 2025 Series C. The platform, built primarily on Anthropic's Claude, serves tens of thousands of lawyers at over 800 firms and in-house teams across 50+ markets including White & Case, Cleary Gottlieb, and Goodwin, with an average annual contract value of $280,000 per client.

HOT TAKE

Nvidia showing up in a legal AI cap table tells you exactly what this is: whoever wins the professional services AI stack will need serious compute, and Nvidia is buying optionality on all of them.

4. Rogo's $160M Series D Is Really a Bet That Wall Street's Junior Class Is Getting Automated

Series D: $160M
Total raised: $300M+
Users: 35,000+
Institutions: 250+

New York's Rogo closed a $160 million Series D led by Kleiner Perkins on April 29, with Sequoia, Thrive Capital, Khosla Ventures, and J.P. Morgan Growth Equity Partners all in the round — bringing total funding to over $300 million. More than 35,000 financial professionals at 250+ institutions, including Rothschild & Co, Jefferies, Lazard, and Nomura, use the platform daily; Rogo's newly launched agent Felix executes multi-step banking workflows autonomously from deal screening through data room diligence. The founders — Princeton classmates who left J.P. Morgan and Lazard jobs in 2022 after GPT-3 convinced them the timing was right — are now backed by the very institutions whose junior analyst pipelines they're restructuring.

HOT TAKE

J.P. Morgan Growth Equity writing a check into the startup automating J.P. Morgan's analyst work is either the most self-aware investment in finance history, or proof that they've already run the numbers on headcount.

5. Big Tech's Brain Drain Just Hit $18.8B — and the Checks Are Getting Bigger

2026 YTD AI lab funding: $18.8B
Ineffable seed: $1.1B
Ineffable valuation: $5.1B

Former DeepMind researcher David Silver — the man behind AlphaGo, AlphaZero, and MuZero — emerged from stealth this week with Ineffable Intelligence and a $1.1 billion seed round co-led by Sequoia and Lightspeed, with Nvidia, Google, DST Global, and the U.K. Sovereign AI Fund on the cap table, at a $5.1 billion valuation. He's the loudest example of a structural shift: according to CNBC, VCs have deployed $18.8 billion in 2026 alone into AI startups founded since January 2025, on pace to surpass the $27.9 billion raised by the prior year's founding cohort. The common thread is researchers leaving frontier labs — where narrow race dynamics have squeezed out whole research areas — to found companies where those bets can actually be made.

HOT TAKE

When the man who taught a machine to beat humans at Go calls his new company "his life's work" and pledges all personal equity proceeds to charity, $1.1 billion at the seed stage starts to feel like it's priced on something other than market comparables.

Most companies learn what their customers think the slow way. They commission a survey, wait six weeks, pay $80,000, and get back data that's already partially stale by the time it influences a decision.

Aaru, founded in March 2024 by Cameron Fink, Ned Koh, and John Kessler, does something fundamentally different: it generates thousands o…

Read more

The Mental Model: Platform vs. Product

Kleiner Perkins partner Mamoon Hamid didn't call Rogo "a great AI tool for banking." He called it "the operating system for an entire industry." That distinction is the Platform vs. Product mental model in one sentence.

A product solves a specific workflow. A platform becomes the layer that other workflows depend on. The shift from one to the other usually happens quietly, then suddenly. One day you're an AI that helps analysts pull comps faster. The next day, 250 institutions have embedded you into their origination process, their pitch decks, their CRM data, and their deal screening. Now removing you is a rip-and-replace project.

Rogo's forward-deployed bankers, experienced finance professionals who work directly inside partner institutions, are the physical manifestation of that strategy. They don't just sell the tool. They embed it into how each client firm operates, from analyst to managing director. That's not customer success. That's platform lock-in built through human relationships.

This is the same pattern we see in every category where a platform wins: the value isn't in the software, it's in how deeply the software is woven into existing workflows. Once it's woven in, the cost to remove it is higher than the cost to keep it.

The Contrarian Take

Here's what I'd watch for: Rogo's biggest risk isn't Bloomberg or Visible Alpha or any named competitor. It's the same risk every platform faces once it hits critical mass. The institutions that depend on Rogo start asking whether they should build the capability themselves.

J.P. Morgan is literally an investor in this round. That's not just validation. That's a bank keeping optionality on whether to acquire or internalize Rogo's capabilities once they're proven at scale. The best clients often become the most dangerous competitors. Ask any enterprise SaaS founder who's had a Fortune 500 company go from paying customer to building a competing internal tool.

For builders: the platform model compounds in ways a product can't. But it also creates a single point of leverage that large institutions know how to apply. Build the platform. Just know who's watching.

Company Overview

Standard Intelligence is a six-person AI research company headquartered in San Francisco. It closed a $75 million Series A on April 30, 2026, led by Sequoia Capital and Spark Capital, at a pre-money valuation of approximately $425 million. The post-money figure lands near $500 million. That math gives you a company worth half a billion dollars before it has publicly shipped a product, staffed a sales team, or hired its tenth employee.

The investors behind the round are not tourists. Sonya Huang and Mikowai Ashwill represented Sequoia. Yasmin Razavi led for Spark Capital. Angel investors include Andrej Karpathy, the AI researcher who co-founded OpenAI and ran Tesla’s Autopilot program; Stanley Druckenmiller, one of the most respected macro investors alive; and Milan Kovac, former head of Autopilot software at Tesla.

The founding team is Galen Mead (21) and Devansh Pandey (20). They met as teenagers through the Atlas Fellowship, a selective program for high-school students focused on AI alignment. Both concluded, independently and then together, that AGI was arriving faster than institutions understood. Both left their undergraduate programs to act on that conclusion. Mead had been enrolled at the University of Toronto. Sequoia’s written description of them is unusually pointed: “taste, scrappiness, technical courage, and ambition” in a combination the firm found rare enough to write a very large check.

Read more

The Mental Model: Economies of Scale

Economies of Scale is the most boring-sounding model in the book, which is exactly why people underestimate it in banking. Unit costs decrease with volume. At scale, every new customer costs less to serve than the last, because fixed costs, technology, compliance infrastructure, and back-office operations get spread across a larger base.

That £400 million in targeted cost synergies isn't magic. It's two compliance departments becoming one. Two core banking systems eventually becoming one. Two mortgage servicing operations merging. Two sets of branch networks being rationalized. The actual Part VII banking transfer isn't even scheduled until the first half of 2027. The real economics of this deal won't show up until that integration completes.

And that's exactly where these deals go wrong. Santander is pricing in the scale benefits before they've done the hard work of actually achieving them. TSB's previous botched IT migration in 2018 cost it over £330 million and a catastrophic customer service failure. That playbook is sitting in every risk officer's memory at this deal's closing table.

The Contrarian Take

The conventional read is "Santander gets bigger and stronger." The contrarian read is simpler: this deal is a direct admission that organic growth in UK retail banking is functionally impossible right now.

Interest rate margins are compressing. Digital challengers are picking off the most profitable customer segments. Neobanks don't have branch cost structures. Santander couldn't outgrow its way to a 16% ROtE. So it bought it instead.

That's not a strength signal. It's a defensive move dressed in acquisition language. The winners in traditional banking over the next decade aren't going to be the ones who consolidated the most. They're going to be the ones who figured out which costs actually need to exist at all.

Scale economies require you to have the right cost structure first. Doubling down on the wrong one just gives you twice the problem.

Somewhere right now, a first-year investment banking analyst is rebuilding a DCF model at 1 a.m. while their VP sleeps. That analyst costs the firm $200K+ in fully loaded comp, burns out in 18 months, and the model still has a typo on row 47.

Rogo thinks it has a better idea.

On April 29, Rogo closed a $160 million Series D led by Kleiner Perkins. The round included Sequoia, Thrive Capital, Khosla Ventures, and — importantly — J.P. Morgan Growth Equity Partners. Total funding now sits at over $300 million, and the valuation has reportedly hit $2 billion, up from $750 million just three months ago after their Series C.

The numbers behind the business are what make this real. Rogo’s platform is used by more than 35,000 financial professionals across 250-plus institutions — Rothschild, Jefferies, Lazard, Moelis, Nomura. These aren’t pilot customers. These are live, production deployments inside some of the most compliance-sensitive workflows on the planet.

Their newest product, Felix, is an AI agent that executes complex, multi-step financial tasks end-to-end: deal screening, CIM generation, buyer outreach, data room diligence. You email Felix. Felix works the problem. Felix doesn’t need to be managed.

Mental Model: 10x Displacement

Here’s the frame I keep coming back to, from Mental Models: How to Think, Act, and Win: 10x Displacement.

Rogo’s claim is that tasks taking junior banker teams hours now take minutes. If that’s even directionally true at 250 institutions, you’re not looking at a productivity tool. You’re looking at a structural change to who gets hired, when, and how many. Banks don’t shrink their analyst classes out of kindness. They do it when the math changes.

This isn’t hypothetical anymore. J.P. Morgan’s equity stake in the company tells you everything you need to know about whether large institutions are treating this as experiment or infrastructure.

Spence’s Take

I’ve spent a lot of time building in regulated markets at Robinhood across derivatives and prediction markets, and now at /mkt, where we’re tokenizing athlete contracts under Reg A+ with tZERO as our trading infrastructure. One thing I’ve learned: the hardest part of fintech isn’t the product. It’s the compliance surface, the institutional trust, and the pace at which large counterparties actually change behavior.

Rogo solved that problem by going deep with specific clients instead of wide with a generic product. They put “Forward Deployed Bankers” real humans with finance backgrounds inside partner institutions to drive adoption from analyst to MD. That’s not a tech-company move. That’s a services-company move wrapped around a software product. It’s slower. It’s also how you actually win in regulated, relationship-driven industries.

The contrarian take here isn’t that AI is eating Wall Street. It’s that the firms who figure out how to integrate agents into actual institutional workflows not just demos will accumulate such a pronounced efficiency advantage that the laggards won’t be able to hire their way back to parity. The moat isn’t the model. It’s the integrations, the proprietary data, and the institutional trust that takes years to build.

Junior bankers aren’t going away tomorrow. But the class of 2030 is going to look very different from the class of 2020. And Kleiner Perkins just bet $160 million on it.

If this was useful, share it with someone who builds things. And if you want the full toolkit of 50 mental models, my book is coming soon.

- Spencer

Share

Leave a comment

Let's put the number in context before reacting to it.

Anthropic's annual revenue run rate has surpassed $30 billion, up from roughly $9 billion at the end of 2025. That's more than a 3x increase in four months. A large portion of that is driven by Claude Code and Cowork. Customers spending over $100,000 annually have grown 7x in the past year.

So the $900B valuation isn't being applied to a company at $9B ARR. It's being applied to a company that just tripled revenue in a single quarter and is still accelerating.

At $30B ARR, $900B is 30x revenue. That's high, but it's not insane for a company with those growth rates in a category that didn't exist three years ago. If the run rate hits $80-100B by year-end, that multiple compresses fast. Early investors from 2024 are reportedly skipping this round and waiting for the anticipated IPO instead. That tells you something about what they think the public market comp looks like.

The Mental Model: Expected Value

Expected Value is one of the models I write about in the book: Probability × Payoff. Most people see a $900B valuation and filter it through loss aversion. "That's too high, there's too much risk." But Expected Value forces you to do the math, not just the vibe check.

If you believe there's a 40% chance Anthropic reaches a $3T valuation at IPO and a 60% chance it plateaus around $500B, the expected value of the current entry is still above the ask. That's a rough version of the calculation every institutional investor in this round is running. The number isn't "does this feel expensive?" It's "what are the probability-weighted outcomes at exit?"

One institutional investor reportedly prepared to commit $5 billion hasn't even secured a meeting with Anthropic's CFO yet. That's not irrational exuberance. That's a buyer who's already done the EV math and doesn't want to miss the allocation.

The Contrarian Take

The risk everyone's avoiding talking about is infrastructure dependency. Amazon has committed up to $25 billion with 5 gigawatts of compute, Google has pledged up to $40 billion with another 5 gigawatts through a Broadcom joint arrangement. Anthropic's growth is real. Its dependence on two hyperscalers for the compute to sustain that growth is also real.

At $900B, you're paying for the revenue trajectory. You're also implicitly betting that Amazon and Google's infrastructure commitments hold, that no new model architecture makes today's compute-intensive training obsolete, and that enterprise customers keep converting at the current rate.

Those are good bets. They're not guaranteed bets.

The investors who get this right will have done the Expected Value math clearly, including the downside scenarios. The ones who get it wrong will have looked at the revenue number, skipped the denominator, and bought the story.

There's a version of this for every raise at every stage, including ours at /mkt. The discipline is the same: know your probability inputs before you commit to a payoff.

1. The AlphaGo Guy Just Raised $1.1B — Before Writing a Single Line of Product Code

Read more

Three years ago, Plata gave credit cards to people who'd never had one. Now it's the most valuable privately held digital bank in Latin America.

This month, Plata closed a $405 million Series C at a $5 billion valuation, led by Bicycle Capital with new backers including Qatar Investment Authority, BTG Pactual, and a slate of U.S. university endowments. The company surpassed $600 million in annualized revenue and built an $800 million loan portfolio, all in under three years. Its CFO says Plata reached that revenue milestone faster than any digital bank in history.

That last claim is doing a lot of work. Let's stress-test it.

Nubank, the Brazilian neobank and the most valuable fintech in Latin America by market cap, took roughly five years to cross $600 million in annualized revenue. Plata did it in three. The whole technology stack, including core banking, CRM, and the AI credit risk engine, was built in-house by a team of over 800 STEM professionals. That's not a feature. That's the business.

Here's what makes Plata genuinely interesting: more than 40% of its 3.5 million active customers came through referrals and organic channels, and over 750,000 of them received their first-ever credit card through Plata. That's not a marketing story. That's product-market fit in a market where incumbents failed to show up.

The Mental Model: Disruptive Innovation

Clayton Christensen's framework is one of the most misused in tech, but Plata is a textbook application of it.

Disruptive Innovation works like this: you enter at the bottom of a market where incumbents don't bother because the economics look bad. You build for the customer everyone else ignores. Then, once you've established trust, distribution, and infrastructure, you move upmarket into the profitable segments the incumbents care about.

That's exactly what Plata did. Mexican banks didn't aggressively serve first-time credit customers. The underwriting was hard. The economics looked thin. Plata went there anyway, built proprietary AI to make the underwriting work, and used those 3.5 million relationships as the foundation to launch full deposit banking in March 2026.

That pivot from credit cards to full banking, and into regulated banking infrastructure with deposit and debit capabilities, is exactly the move that converts a fintech into something incumbents can no longer ignore. They're not competing with Nubank now. They're competing with Banamex, BBVA Mexico, and Santander.

The Contrarian Take

Everyone covering this round is calling it a fintech story. It's actually a regulated banking story, and that's where it gets harder.

Credit-led growth is forgiving. You issue cards, you optimize for approval rates and charge-offs, and you iterate. Deposit banking is different. Now Plata is a full bank, which means capital requirements, reserve ratios, regulatory examination cycles, and all the infrastructure costs that come with holding deposits. The growth metrics that got them to a $5 billion valuation were built on a lighter-weight model. The next chapter requires a different kind of operational discipline.

I've spent the last 6 months building /mkt inside the regulated fintech stack in the U.S. The regulatory wrapper around a securities platform is the same class of problem: it doesn't kill companies, but it changes the pace, the cost structure, and what "winning" looks like. Plata's raise is genuinely impressive. Whether the $5 billion holds depends entirely on how they perform as an actual bank, not just as a credit product.

The difference between a great fintech and a great bank isn't the app. It's the balance sheet.

Share

Leave a comment

A startup with no product, no revenue, and no published roadmap just raised $1.1 billion at a $5.1 billion valuation. At the seed stage.

That’s not a typo.

Ineffable Intelligence, a London based AI lab, emerged from stealth yesterday with the largest seed round ever raised in Europe. The round was co-led by Sequoia and Lightspeed, with participation from Nvidia, Google, DST Global, Index Ventures, EQT, and the UK government’s own Sovereign AI Fund.

The founder is David Silver, the former lead of DeepMind’s reinforcement learning team and a UCL professor. He’s the person behind AlphaGo, the system that beat the world’s best Go players in 2016 and genuinely changed how researchers think about machine intelligence.

His thesis for Ineffable is a direct challenge to the current AI consensus. Rather than training on human-generated text like most large language models, Ineffable is building a “superlearner” an AI system that discovers knowledge through trial and error, without relying on human data at all. TechCrunch Silver’s stated mission: “make first contact with superintelligence.”

There’s one more detail worth flagging. Silver has publicly committed to donating 100% of his personal equity gains to high-impact charities via the Founders Pledge. He’s not doing this for the exit.

Mental Model: Betting on the Jockey, Not the Horse

In venture capital, there’s a long-running debate: do you back the idea or the person? Most experienced investors will tell you it’s almost always the person, especially at the earliest stages when there’s no product to evaluate.

Ineffable Intelligence is that model operating at full intensity. The company has yet to release a product, generate any revenue, or publish a roadmap for what it intends to build. Finance Sequoia and Lightspeed didn’t fund a deck. They funded David Silver’s track record, his scientific worldview, and their conviction that whoever solves reinforcement learning at scale builds something no LLM can replicate.

The horse doesn’t exist yet. The jockey has won every race he’s entered.

This model cuts both ways. Betting on the jockey works when the jockey is genuinely singular someone with a track record that’s not just strong but demonstrably different from everyone else in the field. It fails when investors confuse pedigree with results, or when the market context changes faster than the founder can adapt.

Silver’s edge is that reinforcement learning at scale is his specific area. He’s not pivoting into AI from something else. He built AlphaGo. He built AlphaZero. This is his career’s work, and now he’s betting the rest of his career and his entire financial upside on proving it can go further than any LLM.

That’s a different kind of founder signal.

Spencer’s Take

Here’s the contrarian read nobody’s saying out loud: this round is also a bet against OpenAI’s architecture.

Silver’s whole premise is that scaling LLMs on human data hits a ceiling and that the next leap comes from machines learning from experience, not from ingesting more of the internet. His argument is that creating bigger chatbots via LLMs is a dead end, and that whoever cracks scalable “experience loops” gets a compounding platform that isn’t bottlenecked by human knowledge. Pathfounders

If he’s right, OpenAI’s $852 billion valuation is built on the wrong foundation.

If he’s wrong, $1.1 billion becomes a very expensive PhD dissertation.

I’ve spent enough time building in regulated, capital-intensive markets from derivatives infrastructure at Robinhood to athlete tokenization at /mkt to know that the most important bets often look ridiculous before they look obvious. The LLM consensus is real, well-funded, and deeply entrenched. Those are exactly the conditions that produce the biggest blind spots.

I’m not saying Silver’s right. I’m saying the investors who backed him aren’t idiots, and the rest of us should probably understand why they did it before dismissing it.

Watch the jockey.

Athletes have always been the product. Now they’re buying the platform.

Earlier this month, Factory Holdings, a family office built for athletes, entertainers, and next-gen wealth creators led a $4.7 million investment into Kalshi’s Series E financing round, bringing together 74 investors across sports, entertainment, and business. PR Newswire The names on the cap table read like an All Star ballot: Marshawn Lynch, Richard Sherman, Diana Taurasi, Nneka Ogwumike, Breanna Stewart, Connor McDavid PR Newswire, and more. They came in alongside a16z’s Cultural Leadership Fund.

The round valued Kalshi at approximately $11 billion and included participation from Andreessen Horowitz, Sequoia Capital, CapitalG, ARK Invest, Charles Schwab, and Y Combinator. Yahoo!

This isn’t a celebrity endorsement play. Kalshi is the infrastructure layer for how real-world events get priced and traded. And athletes just bought a seat at that table.

Factory’s mission, according to founder and CEO Keenan Beasley, is straightforward: the people who shape culture should also own the systems they power. PR Newswire

Hard to argue with that.

Mental Model: First-Order vs. Second-Order Thinking

Most people look at this deal and think: athletes investing in a hot fintech company. That’s first-order thinking. It’s accurate, but it’s shallow.

If 74 athletes build fluency with prediction market infrastructure, how contracts are structured, how liquidity works, how regulated markets function, they become dramatically more sophisticated capital allocators. That changes who they trust with their money. It changes what deals they’ll evaluate. It changes what platforms they’ll eventually build or back.

The second-order effect of this investment isn’t just a return on Kalshi. It’s a cohort of high profile, high-influence individuals who now understand financial infrastructure from the inside. That’s a fundamentally different kind of athlete than the one who signs a brand deal and moves on.

Spencer’s Take

I’ve been building /mkt since December 2025, a platform that lets athletes tokenize their brand and earning potential using Reg A+ offerings with tZERO as trading infrastructure. The question I get most often is: are athletes ready for this?

The Kalshi deal is my answer.

Seventy four athletes didn’t accidentally end up on the same cap table. Factory organized them. That’s the important part. The infrastructure to aggregate athlete capital to make collective ownership possible without requiring each individual to have a family office, a GP relationship, and years of deal flow is being built right now.

That’s exactly the gap /mkt was built to address on the other side: the fan to athlete economic relationship. Fans want access to the upside of the athletes they follow. Athletes want to convert influence into durable ownership. Regulated markets are where those two things meet cleanly.

What Factory did with Kalshi is proof that athletes aren’t passive participants in the financial economy anymore. They’re learning to use the instruments. And when athletes understand how regulated markets work, every product built in that space gets a more sophisticated, more engaged set of potential partners.

The ownership era isn’t coming. It’s already here. The question for builders is whether you’re designing for the athletes who are still waiting to be told what to do with their money or the ones who are already on cap tables and asking better questions.

Build for the second group.

Read more

They just started letting businesses settle in stablecoins again.

That’s the signal.

After years of skepticism post-2022, one of the largest payments companies in the world is re-engaging with crypto rails. Not trading. Not speculation. Settlement.

Different game.

Stripe processes over $1 trillion in annual payment volume. When they touch something, it’s not experimental. It’s because customers are already asking for it.

And they are.

1. What’s actually happening

Stablecoins aren’t new. But their usage is.

• Over $10 trillion in stablecoin transaction volume annually

• USDC and USDT dominating cross-border flows

• Emerging markets driving real adoption, not speculation

This isn’t about buying tokens. It’s about moving money.

Faster. Cheaper. 24/7.

If you’re a business in Nigeria, Argentina, or Turkey, stablecoins solve a real problem: access to dollars without relying on fragile local rails.

Stripe stepping back in means demand has crossed a threshold.

Not hype. Utility.

2. Mental model: Infrastructure Inversion

New technologies don’t replace systems overnight. They start at the edges.

Then they move inward.

Call it Infrastructure Inversion.

At first, stablecoins were used by traders and crypto-native companies. High risk tolerance. Low regulation.

Now they’re moving into:

• Cross-border payments

• Treasury management

• Platform payouts

Same tech. Different users.

And eventually, they don’t look like “crypto” anymore. They just look like better infrastructure.

Why this matters now

Timing.

Two things changed:

1. Regulation is getting clearer in key markets

2. Stablecoins proved they work at scale

That second point matters more.

You can argue about crypto all day. But you can’t ignore $10T in volume.

Markets don’t care about narratives. They care about function.

Stripe isn’t early here. They’re late on purpose.

They waited for the system to stabilize, then re-entered when the risk-reward made sense.

That’s how serious companies operate.

What founders are missing

Most founders still treat stablecoins like a feature.

“Let’s add crypto payments.”

That’s shallow thinking.

The real shift is backend:

• How money moves between countries

• How fast settlement happens

• Who controls the rails

If you’re building anything with payments, this changes your cost structure and your margins.

Ignore it, and someone else undercuts you.

Spence’s take

A lot of people still think crypto is dead or irrelevant.

That’s lazy thinking.

Speculation cooled off. Infrastructure didn’t.

And infrastructure is what compounds.

Here’s the contrarian angle:

The biggest winners in crypto won’t look like crypto companies.

They’ll look like boring fintech products with better margins and faster rails.

Stripe sees that.

Most founders don’t. Yet.

No investment advice. Just watching where the rails are moving.

Let that sink in.

Not total funding. Not acquisition price. Individual talent packages.

In one case, offers are rumored to exceed $100 million over multiple years for a single researcher. At the same time, competitors like OpenAI, Google DeepMind, and Anthropic are in active bidding wars for the same small pool of people.

This isn’t normal hiring.

It’s price discovery for intelligence.

The numbers behind the frenzy

AI funding is already at historic levels.

• ~$300 billion in global venture funding in Q1 2026

• Over 60% concentrated in a handful of AI companies

• Frontier labs raising tens of billions per round

Now layer on talent.

There are maybe a few hundred researchers globally who can materially move frontier models forward. Not thousands. Not even close.

So when capital floods into AI, it doesn’t spread evenly.

It hits a bottleneck.

And that bottleneck is people.

Mental model: Supply and Demand

This is the simplest model. It’s also the most useful here.

Qd=Qs at equilibrium priceQ_d = Q_s \text{ at equilibrium price}Qd​=Qs​ at equilibrium price

Demand for elite AI talent has exploded.

Supply has barely moved.

Read more

After years of private market hype, the Swedish fintech filed publicly and is now targeting a valuation around $6.7–$7.5 billion. That’s down sharply from its $45.6 billion peak in 2021.

Same company. Same product. Completely different market.

And it’s not alone.

Across 2026, multiple high-profile companies have either delayed IPOs, downsized offerings, or accepted lower valuations to get public. Market volatility and rate uncertainty are forcing discipline back into pricing. Public investors aren’t buying growth at any price anymore.

That’s the shift.

The data behind the reset

Let’s zoom out.

• Global venture funding just hit ~$300 billion in Q1 2026

• IPO activity is still inconsistent, with many companies waiting on the sidelines

• Late-stage valuations from 2021 are still working their way down

There’s a mismatch.

Private markets moved fast on the way up. Public markets are moving slower on the way down.

That gap has to close somewhere.

Right now, it’s closing at the IPO.

Mental model: Market Clearing Price

Every market has a clearing price. The point where buyers and sellers actually transact.

Not where founders want it. Not where VCs marked it last round.

Where real demand shows up.

Klarna is a clean example.

P∗ is where supply equals demandP^* \text{ is where supply equals demand}P∗ is where supply equals demand

For years, private markets operated above that equilibrium. Cheap capital, aggressive growth assumptions, and low rates pushed valuations higher.

Public markets are less forgiving.

They force price discovery in real time.

When Klarna goes out at a fraction of its peak valuation, that’s not failure. That’s the market clearing.

What founders get wrong

Too many founders anchor to their last round.

That number becomes identity. It drives hiring, burn, and expectations.

But it’s not real until someone else pays it.

In this market, three things matter more than your last valuation:

1. Revenue quality

2. Path to profitability

3. Predictability

Growth still matters. But it’s being repriced.

The companies getting out now understand that. They’re optimizing for certainty over ego.

Why this creates opportunity

Resets feel painful. They’re also necessary.

When prices correct, two things happen:

• Strong companies become investable again

• Weak companies can’t hide behind inflated marks

That’s how markets clean themselves up.

It also opens the door for new entrants.

If you’re starting today, you’re not competing with 2021 expectations. You’re building in a market that values efficiency and real economics.

That’s a better game.

Spence’s take

The narrative says the IPO window is closed.

That’s wrong.

It’s open for companies willing to accept the price the market gives them.

That’s a harder pill to swallow than waiting.

But here’s the contrarian angle:

The companies that go out now, reset, and rebuild credibility will have an advantage.

They’ll be priced for reality.

And in the long run, reality compounds better than hype.

No investment advice. Just how markets work.