The Setter 30 dropped. The better question: which AI startup should you join?

A list went around this week — “the Setter 30” — ranking the private AI companies investors are most eager to buy into. It's a useful snapshot of where the smart money is looking. But it quietly answers the wrong question for most of the people sharing it.

The list tells you what investors want to own. The question nearly everyone reading it actually has is different: where should I go work? Those two questions have surprisingly different answers, and confusing them is how talented people end up joining the wrong company for the right-sounding reason.

Why “great to own” and “great to join” diverge

An investor buys small slices of thirty companies and only needs a handful to win big; the losers are capped at the cheque they wrote. Their job is to own a basket. You get one career bet at a time, paid in years of your life and a slug of equity you can't diversify. A name that's a fine line item in a fund can be a mediocre place to spend the next three years of yours.

An investor needs a basket that wins on average. You need the one seat that's right for you.

The valuation trap

The household-name frontier labs are the obvious draw. The work is real and the logo is gold on a CV. But many now carry valuations in the $300B–$800B range. Here's the uncomfortable maths: your equity upside is roughly “where it ends up” divided by “where you joined.” Join at $800B and, for your shares to triple, the company has to become a $2.4 trillion business. Not impossible — but a lot of the easy growth is already priced in. You're buying prestige at close to full price.

Where the more interesting bet often sits

One widely-shared view — and I think it's broadly right — is that the sharper career bet is often the vertical AI company in the $10B–$30B range: a business applying AI hard to one industry. The names people point to:

• Harvey — AI for law.
• OpenEvidence — AI for medicine.
• ElevenLabs — AI voice.
• Cursor — AI for writing code.
• Replit — AI for building software.

Big enough to be real — actual customers, actual revenue — but small enough that the equity can still multiply several times over. By contrast, AI-flavoured fintech tends to be safer but with a lower ceiling. None of this is investment advice; it's a way to think about the trade between prestige, upside and risk.

But upside isn't the only axis — and arguably not the main one

If you optimise purely for equity, you'll talk yourself into a bad fit. Three things matter at least as much as the valuation:

1. What will you learn, and who from? Three years next to people who are better than you compounds harder than any option grant.
2. Does the company actually reach customers? Real revenue beats real funding. A company that sells has a future you can plan around; a company that only raises has a countdown.
3. Can you live with the risk? “Safer” is not “worse” if it fits the life you want. Don't let a leaderboard shame you into volatility you'll resent.

A simple way to decide

Score each option you're seriously considering, 1–5, on four axes: learning, people, upside-to-valuation, and risk you can stomach. Weight them to your own situation — early career leans learning and people; later career can take a bigger upside swing. Then do the thing the list can't do for you: pick the seat, not the logo your relatives will recognise.

There is no ranking that knows your circumstances. The right company is the intersection of where AI is genuinely heading and what you, specifically, can become excellent at. A list is a starting point for that conversation — never the end of it.