The Short, Sweet Life of Clawdbot

The short, sweet life of Clawdbot is over. On Saturday February 14th, Peter Steinberger announced that he’s joining OpenAI. The Clawdbot => OpenClaw saga—Part 1 at least—has run its course.

I’m not trying to be overly dramatic, and I’m not saying this is a bad thing. The opposite, in fact: something had to give, and if you want a Clawdbot-like set of capabilities to scale, to become reasonably and widely available (I do), with a shot at remaining a viable open source entity (this too), Steinberger landing at OpenAI in the way he’s doing it is pretty close to an ideal outcome.

Unsustainable, Yet …

Why did something have to give? If you listen to his 3 hour 19 minute interview with Lex Fridman (recommended), it’s clear that Steinberger’s situation was unsustainable. He was losing money keeping Clawdbot => OpenClaw alive and growing. His quality of life was suffering. He was being stalked by crypto-bros, VCs (near neighbors!), and the press. The situation got bad enough that he considered just walking away at several points.

Yet all the while this insane and wonderful thing was emerging: Steinberger had connected up the “Opus 4.5 moment” frontier AI models in ways that nobody else had thought to, and OMG did it hit a sweet spot. Record-setting, nerds-and-normies adoption; a virtuous cycle of adoption-fueled and community-led expansion that exposed new emergent AI possibilities. The momentum gained in such a short time was unprecedented. Steinberger didn’t want to kill this off, didn’t want to leave the community high and dry.

Ideal-ish Outcome

Why do I think this outcome was close to ideal? Steinberger isn’t just being acqui-hired; he and OpenAI have apparently agreed to keep OpenClaw open source under the umbrella of a new foundation, much like Anthropic did when it donated the Model Context Protocol to the Agentic AI Foundation, and provide funding for the new entity. This, combined with the meteoric uptake Clawdbot => OpenClaw has seen, gives the project a decent shot at life. There are no guarantees in open source—people vote with their feet— but this is as good a shot as a very young project can hope to have.

Based on my 3 hours and 19 minutes with Peter and Lex, I came away really liking Steinberger and I’m fully aligned with his positions on a number of topics, including:

I don’t do this for the money. I don’t give a f—.

He could have cashed out in 100 different ways, the most obvious being creating a VC-funded, OpenClaw-focused startup. But he opted for a course that will almost certainly put less money in his pocket (though plenty, I’d bet!), but will also give the OpenClaw project and community a good shot at thriving.

Winners and Losers

OpenAI gets the big win here; they get to bask in the momentum and good vibes of a project named for their arch-competitor, and they capture arguably the best person in the world to bring a Clawdbot-like offering to the OpenAI ecosystem.

We—normies and geeks alike—also win.

Steinberger mentions that he was also talking to Meta. OpenAI is much better for us: they can execute and are a couple steps lower on the bad-guys ladder. Count our blessings that the Musk-hole wasn’t in the running.
Foundation plus OpenClaw remains open source with deep-pockets funding. Win.

Finally, Steinberger wins. I’m happy for him. He connected the dots in a new and surprising way, he worked really hard, he kept to his principles. I’m happy he is able to get what he really wants as a result.

He and I share some fairly specific background: experience building and exiting startups, success in the Apple iOS ecosystem, being drawn back into coding and building things by the AI emergence, amazement and fascination (as opposed to fear and angst) at what is emerging. So, good for him, I’m really happy for him and pleased he didn’t just cash out.

The loser? Anthropic. Guys, maybe you avoided the obvious mistake of sending Steinberger a lawyer letter around the Clawdbot name issue, but you still managed to deeply alienate him. This amazing Clawdbot thing was named after your great Claude model, a model that exemplifies the April-to-November 2025 model revolution, which made Clawdbot possible. How do you manage to blow this? How do you come out looking like the bad guys? I know you aren’t; still love you; but maybe fire whoever was giving you legal and PR advice on this one.