Our 2026 LP Meeting

Our 2026 LP Meeting

Plus more data deals and NYC’s ballot measures

Yoni Rechtman

Oct 30, 2026

9

1

Share

On Wednesday, we hosted our annual meeting. We had our ≈25 biggest institutional investors, representing ≈90% of our capital, for one day.

The bulk of the meeting was a presentation that my three partners and I gave:

Will presented our view on the macro environment.

Sam presented our view on AI as a reflection of the macro.

I presented how we’re investing now as a reflection of our view on AI.

Kevin presented the business of the firm.

I won’t share the full slides (they won’t make much sense without the live talk track), but I did want to share some version of the thinking.

TL;DR: neutral to short AI beta. Long the alpha.

What we’re long:

Selling outcomes.

Software products, not software business models. Selling / verticalizing around work means bigger markets and better GTMs (no one wants to buy software).

Second order effects of AI.

Our markets and society are based on the presumption of effort being more expensive than attention. AI very clearly flips that.

The edges of the AI supply chain.

Bottlenecks, physical assets, and data that have convex upside on diffusion.

Post-AI builders.

The “of the moment” founders are technical, commercial, and rangy. AI increases variance and this is especially true in talent markets.

What we pass on:

Infinite capital consumers.

Anything that needs to access infinite money to work / where our money doesn’t matter (we don’t have infinite money!)

Paying for legibility.

Self explanatory. Becomes a winning game, not a picking game.

Raw capability

. “Software can do a lot of cool stuff” is not a view on how you uniquely capture that value yourself.

Trades dressed as investments.

Trading is a great business and increasingly there is actually liquidity to get out but it’s not our business or what we think we do well.

Selling outcomes.

Outcomes always required intelligence, and intelligence lived in people. People produced outcomes using software. Now software can use itself. Normal companies spend ≈3% of revenue on software vs ≈25% on labor and ≈10% on services, and no one has ever wanted to buy what SaaS companies sell. Software can do work now, and

that’s easier to sell than workflow

, into much bigger markets.

Sometimes that’s a product company delivering a service (vertically integrating around software, using FDE to own the outcomes around software, or verticalizing around hardware). Sometimes it’s buying the business, then transforming it via a GBO or AI rollup. If you sufficiently transform the earning power, keep the upside.

Modeling software after SaaS

Our buyout thesis

Second order effects of AI.

When effort is free, systems built on effort break. AI won’t make more attention or trust, but it creates infinite demand for both. This is happening/will happen across markets/transactions based on security, trust, attention, scarce physical assets, the law, etc. So much of our economy and society relies on low agency to function but now agency is infinite.

We back the companies that control or rebuild those systems

.

Do you know a guy?

Investing in the second order effects

The edges of the AI supply chain and physical AI.

Models and compute are hyper-legible and hyper-capitalized, creating a proxy war between the capital markets megaliths. Instead we want convex opportunities

where our money actually matters

. Pricing power sits at the constrained/narrow points through which the (physical) AI supply chain has to run. We’ve been focused on owning the business that models/robotics/hardware makes possible, not the hardware in line with our view on software (outcomes win). Finally, data and its value-added byproducts are emerging as a critical input to the supply chain and a new part of its capital structure.

Data-native capital is coming

Post-AI builders.

We’re very excited about the new emerging models for both business building and founder archetypes. The companies getting started today are getting built with really different assumptions about the future than businesses of even just a few years ago. That means that, irrespective of age or prior experience, we are looking for “post-AI” founders. These founders are super competitive, rangy, commercial, and technical, and they’re building with a deep belief about the inherent worthlessness of application software.

Coherence is everything: one choice/one POV on tech, GTM, org chart, and talent. Any company you build has to be a reflection of that bet, expressed through every choice.

There will only be four jobs

Cargoculting and coherence

Being Business Nerds

We’ve always been business nerds. Every new technology makes new business models and capital structures possible, and we want to be first to them.

This was my favorite slide

We want to be

maximally creative, entrepreneurial, and open-minded in thinking about business building and terminal value

. This is also a really important part of how we attract and screen founders. There’s a certain type of person (some of whom are great!) who’s just not interested in these questions and probably will not value partnering with us / will never take our money. But for our kind of founder, this type of thinking is at least interesting and hopefully actively helpful to assemble around themselves.

By foregrounding this approach,

we can select for the founders that we want to back

, who are themselves maximally creative and entrepreneurial.

The biggest risks to our worldview

If we’re wrong, we think we know why:

Maybe the Great Sorting is fully predetermined. There’s no crossover, good businesses stay just good businesses forever, and getting in early buys us nothing / we have no shot at stuff that matters without paying up.

Maybe the infinite money is truly infinite: capital asymptotically concentrates in infinite TAM, which ironically makes it harder, not easier, for almost everyone to raise, especially Series A/B rounds.

Maybe there are no unloved gems and the market is more efficient than we think. We get our price by backing founders who are not optimizing for fundraising (definitionally, low price/high CoC) and ideas that are hard to grok. That puts us further out on a limb than anyone else, and if the market is right about them, we eat it.

And if RSI solves scarcity, labs and leisure end up as the only businesses with any value. Or if P(doom) is way higher than we think, we’re all dead. Either way, there’s no more money so who cares!

Diseconomies of Scale in Data Markets

Will data value ultimately scale with novelty (IP) or volume (oil)?

Small businesses ($1-2M sales) will routinely sell company data for 10-50% of revenue. Sometimes even >100%!

Spirit airlines ($5B sales) sold its data for $10M (0.2% of revenue)

No, data values and revenue should not be perfect correlates but you’d expect it to be a little closer than this!

There’s a few possible explanations for this divergence:

There’s very few buyers TODAY for truly scaled data which suppresses prices TODAY. And even fewer individuals within those buyers with the budget and centralized procurement to actually deploy big dollars. So Google just got a great deal on Spirit because the market is so immature.

The bankruptcy process was not well run/not made for this type of asset (no samples). It was a firesale so don’t extrapolate from firesale prices.

There’s diminishing marginal returns on the data, trajectories, etc. beyond a particular size. You don’t get 10x the benefit at 10x the scale. At least for now, post training and RL require novelty not size.

There’s increasing transaction costs that suppress the useful value. At real scale, getting the data out and making it usable (no PII, full corpus, etc) is just too much of a pain. It’s easier to work with smaller sets.

At scale you’re competing more with synthetic data. Real world data is necessary to establish a baseline but not to fully populate or scale an environment.

#1 and #2 are potentially bullish because they’re transitory. #3 is bearish. #4 is neutral and bullish for infrastructure companies. #5 is probably right on a long enough time horizon.

Whether or not the diseconomies of scale in data sales/values are persistent is going to be among the most important questions in the data market and toward the financialization of data and the emergence of robust capital markets around it.

I would love to talk to anyone who has worked on larger data deals: $5M+ transactions. Hit me up!

Data Deals

The obvious next step for me as I think about data-native capital and data-native companies is to work on some deals. I’m interested in helping someone sell data/acting as a broker or advisor on a couple deals. Seems like the natural and necessary step to learn more here.

NY Ballot Measures

For my NYC people - I am voting “yes” on all five ballot measure.

From my friends at

Abundance NY

(read their whole voter guide):

Those are the Commission on Government Efficiency five proposed amendments to the City Charter—essentially NYC’s constitution—to simplify and modernize a few outdated elements of local government that can only be improved via ballot referendum.

Put simply, they will:

Make it faster for new homes and outdoor dining to be permitted, and for street safety and accessibility improvements to be implemented.

Make it easier for small businesses and nonprofits to do business with the city, and for the city to lease office space for its operations.

Set targets for savings into the city’s Rainy Day Fund.

9

1

Share