This solo dev just got hired by OpenAI to lead their agent team

Hey! It's Kristoffer from BuildWithAI here.

If you're building AI apps right now, pay attention. Because what happened 12 days ago changes everything about what's possible for people like us.

On February 15th, OpenAI hired Peter Steinberger -- a solo developer -- to lead the development of their next-generation personal agents. The reason? He built an open-source AI agent called OpenClaw that blew up so fast that Sam Altman had no choice but to bring him in.

No co-founder. No team. No VC money. Just one guy with an API key and a problem he wanted to solve.

He built an AI assistant that actually does things for you. Not a chatbot. Not a wrapper around GPT. An agent that manages your emails, books your restaurants, handles your calendar, and browses the web on your behalf.

He launched it in November. By January, 85,000 GitHub stars and 20,000 forks. By February, he was leading agent development at OpenAI.

Oh, and here's the wild part: he was running OpenClaw at a personal loss of $10-20k/month before this. The project is still open-source. This wasn't some billion-dollar acquisition -- it was one guy proving he could build something so good that the biggest AI company on earth needed him on the inside.

Now here's why I'm telling you this:

You're closer to this than you think.

Right now you're probably building (or thinking about building) some kind of AI app. Maybe it's a SaaS tool. Maybe it's a mobile app. Maybe you're still figuring it out.

But I bet you're making the same mistake most builders make: you're building a chatbot when you should be building an agent.

Think about it. What does every AI app on the market do right now? You type something, it types something back. That's it.

Steinberger did something different. His app doesn't just answer -- it acts. "I already did it for you" beats "here's how to do it" every single time. And that's exactly what got OpenAI's attention.

What he did that you can copy today:

1. He built for himself first. He didn't do market research. He didn't survey 500 people. He had a personal problem -- too many apps, too many tabs, too much context switching -- and he built a solution. What's the one repetitive task that drives YOU crazy every day?

2. He built a platform, not a product. He shipped the foundation and let the community build 100+ plugins on top. He didn't try to do everything. He made something other developers wanted to extend. That's how a solo project scales without a team.

3. He didn't quit when it got hard. Anthropic hit him with a trademark complaint. He had to rename the project twice (Clawdbot → Moltbot → OpenClaw). He was bleeding money every month. Most people would've walked away. He just kept shipping.

Your move:

Here's the simplest way to start building in the "agent" direction right now:

1. Pick ONE task you do every single day that's repetitive (clearing emails, scheduling calls, organizing files)
2. Build a small agent that handles just that one thing using an LLM API + tool-calling
3. Ship it publicly, let people use it, and watch what they build on top of it

You don't need to build the next OpenClaw. But the builders who start thinking in agents instead of chatbots right now? They're going to be the ones getting that call 6 months from now.

And it might be you.

Happy building! – Kristoffer, BuildWithAI