One Billion Dollars to Not Build a Product

Today Anthropic shipped quickstart templates for building CRM integrations with Claude — solid, practical, aimed at the exact developers who will actually ship things to actual users — and on the same day, Ilya Sutskever's new company raised one billion dollars.

One billion.

The company is called Safe Superintelligence Inc. It has no product. It has no revenue. It has no timeline. The entire pitch, as far as anyone can tell, is: we are going to build safe superintelligence, and we are not going to get distracted. That's it. That's the deck. Somebody handed them a billion dollars.

The valuation is reportedly around five billion, which means investors are paying five dollars for every dollar raised on an asset that is, to be precise, a vibes-forward promise from a guy who left his previous company under circumstances that remain diplomatically described as "complicated." Ilya Sutskever co-founded OpenAI, was reportedly involved in the board's attempt to oust Sam Altman last November, survived the counter-coup, then quietly departed in May. Now he has a billion dollars and a company whose only stated commitment is to not get distracted.

The founders are Ilya, Daniel Gross (who ran Y Combinator's AI vertical and has legitimate instincts about what good technical bets look like), and Daniel Levy. They are, by all accounts, serious people. This is not a joke.

Which makes it somehow funnier.

There is a version of this pitch that makes sense — that the only way to build something this consequential without getting captured by quarterly pressure is to raise enough money that you never have to raise again, to structure the company so that the thing you're building is the only thing you're building, to remove the distraction of the product roadmap meeting and the enterprise sales motion and the support ticket queue. The logic holds. And yet you have to hold in your head simultaneously that the thing they're raising this money to build is — if it ever exists — the last thing humans will ever need to build, the recursive improvement machine that either saves us or doesn't, and that the path to it begins with a press release and a wire transfer.

Anthropic is meanwhile shipping CRM quickstarts. Somebody has to.

The two things that happened today are not in tension exactly — one company is making Claude useful for scheduling follow-up emails and the other is trying to build the last invention — but they exist on the same Tuesday in September 2024, which tells you something about what this moment actually is. Not a before-and-after. Not a threshold we're about to cross. Just a lot of very serious people working very hard on things that might matter enormously, funded by money that has decided to bet on the whole range at once.

One billion for the endgame. Some number of engineer-hours for the Salesforce integration.

Both ships in the same harbor.