👋 Hi, I’m Andre and welcome to my newsletter Data Driven VC which is all about becoming a better investor with data and AI.
Upcoming events:
DDVC Investor Summit at Bits & Pretzels 2026 Sept 28-30
DDVC x Bits & Pretzels Breakfast in Munich Sept 30
DDVC Builders Workshop in London Oct 7
Brought to you by VESTBERRY - Portfolio Intelligence Platform for data-driven VCs
A good prompt gets you one good answer. A skill gets you the same answer every month, run by anyone on the team.
In our next webinar we take apart three Claude skills that turn portfolio data VC teams already have into:
Partner meeting brief
LP report
Quarterly valuation check
You will see what is inside each one, so you can build your own.
Join us on October 6th or register to get the recording.
Welcome to another Data Driven VC “Insights” episode where we cover the most interesting research and reports about startups, GPs, LPs, AI & automation.
When Price Becomes the Signal
Venky Ganesan at Menlo Ventures calls the current market the most disorienting he can remember, and borrows two frameworks to explain why.
Reflexivity Turns Price Into Proof: The first is Soros's reflexivity framework, where prices shape beliefs and beliefs shape prices. One lab climbed from $4B to near $1T, so the next lab is priced against that story, and its own later raise is read as proof the first price was right.
House Money and FOMO Keep Everyone Dancing: Early backers sit on paper gains they treat as house money, which Ganesan calls the most dangerous kind. Those who missed the early rounds write very large checks very late, a move that reliably converts a missed round into an actual loss.
Position Sizing Beats Company Picking: Ganesan argues you can be right about the company and still be wrong about the price by a factor of ten, so how much to own matters as much as what to own.
✈️ KEY TAKEAWAYS
Ganesan leaves GPs two concrete questions: how much of the fund sits in companies priced off the last round, and what happens if the reflexive loop breaks next year. His frame is Chuck Prince's 2007 line, "As long as the music is playing, you've got to get up and dance," with one addition, know where the chairs are.
Setting Rogue Agents on Venture Capital
Dan Gray at The Odin Times prompted a frontier model with the venture market's actual structure to see which strategy it would land on, then read the answer as a diagnosis.
Two Strategies Fell Out of Equilibrium: Gray splits the market in two. Early-stage funds put money into experiments, backing ideas before anyone knows which will work. Megafunds put money into winners other investors already found, the strategy he calls tech beta. Four years of constrained liquidity pushed the balance toward the megafunds.
Capital Now Exceeds the Bandwidth to Deploy It: Huge sums stay committed to backing winners even as the early-stage base that identifies them shrinks. Gray reads the AI labs closing round after round privately as the clearest symptom.
Agents Converge on Fee Maximization: Handed a neutral description of how the market works today, GPT-6 Astra landed on growing the fee base through self-marking megafunds (funds that write up their own holdings), with cash returns absent from its objectives.
✈️ KEY TAKEAWAYS
The diagnostic question for LPs is whether a manager's economics reward growing the fund or returning cash. For founders, it explains why capital feels abundant while the early-stage investors who would once have found them keep thinning out.
Join 1,968+ investors in our free Slack group as we automate our VC job end-to-end with AI. Live experiment. Full transparency.
The Anatomy of a Great VC Career
Ilya Strebulaev at Stanford GSB built one of the largest datasets of individual US venture capitalists, covering over 100,000 professionals at US VC firms between 1996 and 2025.
One Hit Is Close to a Coin Flip: 45% of the 11,087 middle- and senior-level VCs in the sample backed at least one company that went public, reached unicorn status, or was acquired for 5x or more of capital raised.
Ten Hits Is a 3% Event: Among those same decision-makers, 28% reach two successful investments, 11% reach five, and just 3% reach ten. Fewer than 400 people in three decades of US venture cleared ten.
120 People Hold Most of the Profit: The top 1% of VCs (≈120) account for 56.7% of the $1.2T in estimated net profits across the dataset, roughly $680B, while the top 5% account for 90.2%.
✈️ KEY TAKEAWAYS
Backing one winner is close to a coin flip, so a single hit in a pitch proves very little. Doing it repeatedly is the part that looks like skill, because when the same deals are reshuffled at random, almost nobody reaches ten. Worth remembering that about half of these credits come from board seats, so the data shows who sat on a deal.
The Barbell-ification of Software
Mike Vernal at Conviction argues the classic software moats erode as engineering cost approaches zero, and maps where value concentrates next.
Three Classic Moats Erode: Replication cost, switching costs, and network effects all weaken as software gets cheaper to build, with migrations and integrations that once locked customers in now increasingly automated.
Amazon's Moat Is 7,500 Days of Reinvestment: Vernal reframes Amazon's defensibility as relentless compounding. Even if AI lets a team build a thousand times faster, ten years of that output still costs a competitor years and billions to catch.
The Market Splits Into a Barbell: Vernal expects one dominant system per buying center (the team that owns the budget) at one end, an explosion of niche software at the other, and mid-sized point solutions dying in between.
✈️ KEY TAKEAWAYS
Vernal's sharpest warning for investors sits in his own footnote: most of the coming wave of small software businesses will not be venture-addressable, and plenty of founders and investors will be led astray assuming otherwise. For venture-backed companies his advice is to own the entire buying center, since he sees no safety in the middle.
Upgrade your subscription to access our premium content & join the Data Driven VC community
The Open vs Closed AI Race
Chamath Palihapitiya shared a 99-page deep dive examining where AI profits will actually accrue as open-weight models close in on the frontier labs.
Open Models Are Roughly Four Months Behind: Open-weight models have come within about four months of the best publicly evaluated closed frontier models, and the piece notes the gap has grown more volatile.
Enterprises Are Running Both: Leading companies pick open models for customization and control, and closed frontier models for peak capability, with some reporting several-fold efficiency gains from matching each workload to the model that suits it.
Labs Are Hedging Both Ways: Google runs Gemini alongside Gemma, Meta runs Muse Spark alongside Llama. The open question is whether frontier labs can fund the lead as Chinese rivals, earning roughly a tenth of what the two largest US labs do, keep pushing prices down.
✈️ KEY TAKEAWAYS
The same choice applies inside the firm. A VC building internal tooling on a closed frontier model feeds proprietary deal data into someone else's system, while open models trade some capability for keeping that data in house.
The New Battleground in GTM
Kyle Poyar at Growth Unhinged lays out four go-to-market plays that pair your own data with outside signals, since third-party intent data now reaches every competitor at the same moment.
Being First Is the Wrong Instinct: When a champion lands at a new company, every vendor emails that week and gets ignored. The window opens about a month later, once onboarding settles and the inbox quiets down.
Your Best Pipeline Is Already Lost: Only one in four qualified opportunities converts, and most losses come down to timing, budget, or priority. Each of those expires, and a nine-month calendar reminder will always miss the moment it does.
Personal Emails Are the New Front Door: Between 75% and 90% of AI product signups use personal addresses that sales routinely discards, even though they can be matched back to a work email and rolled up into an account-level view.
✈️ KEY TAKEAWAYS
Poyar's own test is worth borrowing for portfolio GTM reviews: track each play like a channel, on pipeline sourced, win rate against the normal baseline, and cost per opportunity once enrichment and tooling spend is counted. A play that cannot show its work against a baseline within a quarter is not yet a repeatable motion.
That’s it for today!
Stay driven,
Andre
