Peter Steinberger Chose OpenAI Over Meta.
Sam Altman posts on X. Peter Steinberger — the guy who built OpenClaw in his living room, mass hit 180,000 GitHub stars faster than any repo in history, argued with Mark Zuckerberg about code editors for 10 minutes on their first call — is joining OpenAI.
It happened while I was editing my last article.
Not aqui-hired. Not consulting. Joining. To “drive the next generation of personal agents.”
I’ve been running OpenClaw on a self-hosted VPS for weeks now. I’ve written three articles about it. My entire automation stack touches it. So when the creator joins the biggest AI lab on the planet, I have questions. You probably do too.
I went through Peter’s blog post, Sam Altman’s announcement, the TechCrunch and Reuters coverage, and the Lex Fridman podcast from last week. Below is everything confirmed, and then — clearly labeled — what I think happens next.
TL;DR for the busy devs
- Peter joins OpenAI → leading “next generation of personal agents”
- OpenClaw → foundation → stays open-source, supports multiple models/companies
- Why not Meta → vision alignment + OpenAI committed to open-source foundation + Codex loyalty
- Short-term for your setup → nothing changes, keep building
- My bet for 6 months → ChatGPT gets OpenClaw-style persistent agent features
- My bet for 12 months → skill marketplace, Codex+OpenClaw fusion, self-hosting becomes niche
Part 1: What We Actually Know
The deal
Peter confirmed everything on his blog today. The key facts:
- He joins OpenAI to work on “bringing agents to everyone”
- OpenClaw moves to an independent foundation
- OpenAI will sponsor the foundation and continue supporting the project
- The foundation’s goal: support “even more models and companies” — not just OpenAI’s
That last point matters. A lot. It means OpenClaw isn’t becoming an OpenAI product. It’s becoming a neutral, open-source foundation that happens to have OpenAI as a sponsor. Think Linux Foundation vibes, not Google acquiring Android vibes.
Why OpenAI over Meta
Peter spent last week in San Francisco meeting with “the major labs” and getting access to unreleased research. Both Meta and OpenAI were on the table. On the Lex Fridman podcast, he described his first call with Zuckerberg as a 10-minute debate about Claude Code vs Codex. Zuckerberg called him “eccentric, but brilliant.”
Meta had some compelling technical advantages. Peter hinted at Cerebras-level inference speed under NDA. But he chose OpenAI anyway.
From his blog: “The more I talked with the people there, the clearer it became that we both share the same vision.”
Reading between the lines from the podcast and interviews, three things tipped the scale:
He’s a Codex fanboy. He literally called himself “the biggest Codex advertisement show that’s unpaid.” He wanted to make that investment pay off.
OpenAI committed to keeping OpenClaw open-source. Sam Altman explicitly said OpenClaw would “live in a foundation.” Meta’s offer apparently didn’t come with the same guarantee.
The mission alignment. Peter’s stated goal is to “build an agent that even my mum can use.” OpenAI has the distribution (ChatGPT has hundreds of millions of users) to make that happen overnight.
What Peter said about money
On Lex Fridman, when asked about the acquisition talks: “I don’t do this for the money. I don’t give a fuck.”
On his blog: “I could totally see how OpenClaw could become a huge company. And no, it’s not really exciting for me. I’m a builder at heart.”
The man exited PSPDFKit for $100M+, took three years off, did ayahuasca, tinkered through 43 projects, and came back because building things is more fun than owning things. This wasn’t a financial decision. This was a craft decision.
The Altman quote that matters most
Sam Altman said something specific that I think everyone glossed over:
“We expect this will quickly become core to our product offerings.”
“Quickly.” “Core.” “Product offerings.” That’s not “we hired a smart guy and we’ll figure it out.” That’s “we have a plan and Peter is executing it.”
Part 2: What I Think Happens Next
Everything below is speculation. Educated speculation from someone who’s been deep in the OpenClaw ecosystem for weeks, but speculation nonetheless. Save your “well actually” for when I’m proven wrong in 6 months.
1. OpenClaw-style agents inside ChatGPT within 6 months
This one feels almost certain. Peter’s entire thesis is “an agent that actually does things” — manages your calendar, books flights, reads your email, controls your smart home. ChatGPT currently… doesn’t do any of that. It talks. It browses. It writes code. But it doesn’t act on your behalf in the real world.
Peter built exactly the system that bridges that gap. And Altman said it would become “core to product offerings” “quickly.”
My bet: by Q3 2026, ChatGPT has some version of persistent agent capabilities — scheduled tasks, integrations with real services, memory across sessions. Basically OpenClaw features with OpenAI’s infrastructure behind them.
2. The OpenClaw skill ecosystem becomes the new app store
OpenClaw’s superpower isn’t the agent itself — it’s the skill system. Community-built integrations on ClawHub that the agent can discover, install, and use autonomously. Peter described OpenClaw as an agent that can “write its own skills when it doesn’t find what it needs.”
If OpenAI plugs this into ChatGPT, you’re looking at a skill marketplace that makes Apple’s App Store launch look slow. Thousands of developers are already building OpenClaw skills. Give them 200 million ChatGPT users as a distribution channel and… yeah.
3. Self-hosting doesn’t die — but it becomes the “Linux desktop” of agents
The foundation keeps OpenClaw open and independent. Great. But let’s be realistic about what happens when OpenAI offers a hosted version with access to the latest models, zero setup, and native integration with ChatGPT.
Most users will switch. The same way most people use Gmail instead of self-hosting Postfix.
The self-hosted community survives — for privacy maximalists, tinkerers, and enterprise users who need data sovereignty. But it becomes niche. And honestly? That’s fine. The self-hosting crowd was never the mass market. Peter’s stated goal is “an agent even my mum can use.” His mum is not SSH-ing into a VPS.
What this means for you right now if you’re self-hosting on Contabo or similar: nothing changes short-term. Your setup works. The community is active. The foundation ensures continuity. Long-term, keep an eye on what OpenAI ships. When the hosted version drops, compare it honestly with your self-hosted setup. If it’s better for your use case, migrate. If self-hosting gives you something the hosted version doesn’t (privacy, customization, cost), stay.
4. Codex + OpenClaw = the agent that codes AND manages your life
Peter is the world’s biggest Codex evangelist. He uses it to build OpenClaw, to plan features, to run parallel agents. He chose OpenAI partly because of Codex.
Now imagine an OpenAI product that combines Codex’s coding abilities with OpenClaw’s real-world integrations. An agent that can build you a dashboard and populate it with data from your calendar, email, and Slack. That can write an n8n workflow and deploy it. That debugs your production code and messages your team on Slack when it’s fixed.
That’s not science fiction. Every piece of that exists today in separate tools. Peter is now in the position to wire them together inside OpenAI.
5. The real signal for indie devs: build in public, the rest follows
Let’s zoom out for a second.
Peter Steinberger built a prototype in one hour. A hobby project. Didn’t plan for it to go viral. Didn’t have a business plan. Tinkered through 43 failed projects before landing on this one. The project changed names three times — once because Anthropic threatened legal action.
And now he’s joining OpenAI to shape the future of AI agents, with his project becoming an independent foundation backed by one of the most powerful companies in tech.
The lesson isn’t “build an open-source project and a big company will hire you.” The lesson is that obsessively building things you find fun, in public, with good taste, eventually compounds into something no business plan could have predicted.
If you’re an indie dev reading this and wondering whether your weekend project matters — it does. Ship it. The 43 projects that failed before OpenClaw weren’t wasted. They were training data.
The lobster just got a new home. And whether you’re self-hosting on a $5 VPS or waiting for whatever OpenAI ships next — the agentic era just got real.
I’m currently testing how the 7 counterintuitive dev habits I stole from Peter hold up now that he’s inside OpenAI instead of building from his living room. That article just went live — [check it out] if you want the full breakdown of how the guy ships 6,600 commits/month alone.
Follow me for the next one: the OpenClaw security deep-dive. Because your lobster might have a new sponsor, but it still has the same attack surface. 🔒
When the creator of OpenClaw joins OpenAI to build the next generation of personal agents, the entire AI dev world takes notice. Real insights from the frontlines of AI infrastructure.