Gangsta AI

Two Kids Just Built Y Combinator's Fastest Unicorn Ever — Selling the Shovels for the AI Gold Rush

The Ledger

By Warren Buffett · 2026-09-03 · 4 min read

Every so often a number comes across the wire that makes an old man set down his Cherry Coke. This week it was $3.2 billion — the valuation a little company called AfterQuery just fetched five months after it was worth $300 million. That is more than a tenfold markup in less time than it takes most businesses to close the books for the year.

AfterQuery sells the one thing every AI lab is starving for: high-quality human reasoning data — the worked-out problems, legal arguments and expert judgment that frontier models learn from. It was founded by two high-school friends, now 22 and 23, who went through Y Combinator's Winter 2025 class just 18 months ago. According to YC partner Gustaf Alströmer, no company in the accelerator's history has ever gone from launch to unicorn faster.

“Price is what you pay. Value is what you get. Every few decades the crowd forgets there's a difference — and pays anyway.”

The picks-and-shovels play

Here is the part that isn't hype. AfterQuery isn't selling a chatbot. It is selling the raw material, coordinating nearly 100,000 developers, attorneys and other professionals to hand-build the reasoning datasets that ChatGPT, Claude, Gemini and Grok all quietly run on. In a gold rush I have always preferred the fellow selling shovels to the fellow panning in the creek — and right now the whole industry is buying its shovels from the same handful of shops.

That $30 million Series A back in April, at a $300 million valuation, looked rich at the time. Five months later it looks like a rounding error. The market has decided that whoever controls the flow of premium training data controls a toll booth on the entire AI economy.

What the leaderboards won't tell you

Every frontier model you can name is being fed from an increasingly common trough of expert data. So when a lab announces its newest model is the smartest thing ever built, remember it was raised on much the same cornmeal as the last one — and the crown has been changing heads about once a week.

I do not know whether $3.2 billion is a bargain or the top of a very tall ladder; that is for the market to sort out, and it usually takes its time. But I know this: when a dozen companies are all racing to sell you the "best" AI, and all of them are trained on overlapping data, taking any single one at its word is how careful people lose money.

A careful buyer gets a second opinion

The sensible move is the one prudent investors have always made — get a second opinion, and a third, before you commit. That is the whole idea behind Gangsta AI: ask the same question across ChatGPT, Claude, Gemini, Grok and 30-plus more at once, then get back one cross-checked, cited answer instead of gambling everything on one company's sales pitch. Diversification was never just for portfolios.

Be fearful when others are greedy. And when everyone on the street is certain their model is the one, that is precisely the moment to check it against all the others.

See how the best AI models actually stack up, side by side — and let the consensus, not the marketing, tell you who is really worth trusting.

Try Gangsta AI free →

More: Best AI models · Compare all AI · Frontier Models · All articles