Peter Steinberger: From PDFs to AI Agents — The Story of the Creator of OpenClaw

A burned-out guy, no desire to code, traveling solo across Europe. Three years earlier he'd sold his company for over €100 million. Now he couldn't even open a code editor.

Three years later, in one hour, from a hotel in Marrakesh with spotty internet, he built what would become one of the fastest-growing repositories in GitHub history.

This is the story of Peter Steinberger. And if you've ever felt like burnout won, like you don't know what to build, or like your best days are behind you — read the whole thing.

The kid who couldn't sit still

Peter is Austrian. In 2004 he enrolled in medical informatics at the Vienna University of Technology (TU Wien). But he wasn't the kind of student who sits in the back row.

While still in school, he launched the university's first Mac and iOS development course. It didn't exist. He created it. And for two years he mentored development teams, teaching other students to build apps for an ecosystem that was just being born.

"I can do it better"

After university, Peter moved to San Francisco. He got a job as a senior iOS engineer at a startup. Everything was normal until he had to display a PDF on an iPad.

It didn't work well. The experience was bad. Any reasonable person's response would have been to complain and move on.

Peter's response was: "I can do it better."

And that's how PSPDFKit was born.

The PDF era: bootstrapping to a billion devices

In 2011, together with Martin Schürrer, Peter co-founded PSPDFKit. The name is terrible — he admits it himself. "I'm really bad at naming things," he's said more than once. Spoiler: this becomes a pattern.

What wasn't terrible was the product.

PSPDFKit started as a side project by two people. No investors. No external funding. Zero. Bootstrapped in the purest sense of the word.

And it grew. Year after year, without raising a dime of outside capital, the company expanded to nearly 50 employees. Their PDF rendering technology ended up running on over a billion devices. Banks, media companies, Fortune 500 firms — everyone used PSPDFKit. It became the gold standard for document handling.

The lesson for founders?

You don't need Silicon Valley to build something massive. Peter was in Vienna. Bootstrapped. No pitch decks, no funding rounds, no convertible notes. Just a product that solved a real problem better than any alternative.

The €100 million exit

In 2021, after a decade without a single euro of external financing, Insight Partners came in with a strategic investment of over €100 million. The first outside money in the company's history.

Peter stepped down as CEO. The company eventually rebranded as Nutrient (at least someone improved the name).

From the outside, it's the perfect story. Bootstrap for ten years, build something that runs on a billion devices, make a multi-million-dollar exit. The end.

But the real story is just getting started.

The dark years: when success destroys you

Thirteen years of high-pressure work took their toll. After the exit, Peter didn't feel liberation. He felt emptiness.

He couldn't code. He didn't want to code. The guy who had built an entire company from scratch, who had taught app development while still a student, who had looked at a broken PDF and said "I can do it better" — that guy no longer felt like opening a terminal.

He bought a one-way plane ticket to Madrid and disappeared.

He traveled. He tried, in his words, to "get his life back." All those years he'd sacrificed building PSPDFKit — the relationships he didn't nurture, the experiences he missed, the versions of himself he never explored.

"If you wake up in the morning and you have nothing that excites you, no real challenge, that gets very boring very fast. And when you're bored, you go looking for other places to stimulate yourself, and that takes you down a very dark path." — Peter Steinberger on the Lex Fridman Podcast

Three years. Three years searching for purpose after having it so clear for thirteen.

Why does this matter?

Because nobody talks about this. In the startup ecosystem — the exit gets glorified. People talk about the before and the moment. Never the after.

Burnout isn't weakness. It's the cost of intensity. And Peter Steinberger, with €100 million in the bank and all the time in the world, couldn't escape it.

The AI wave: 43 failed experiments

At some point in 2025, artificial intelligence brought him back. It wasn't an epic moment of revelation. It was curiosity.

Peter started building. One project. Another. Another. 43 projects before finding the one that mattered.

Some were brilliant in concept. He fed his WhatsApp data into GPT-4.1's massive context window and generated analyses of his friendships so deep that his friends ended up crying when they read them.

He built "vibetunnel," a tool to expose local Mac terminals to the web. Then he used a single AI prompt to translate the entire codebase from TypeScript to Zig. One prompt. The whole codebase.

Every project taught him something. None of them stuck. But he was coding again. And that, after three years of darkness, was already a victory.

Project 44: one hour in Marrakesh that changed everything

November 2025. Peter was on a birthday trip with friends in Marrakesh. And between the chaos of the souks and the mint tea, he spent one hour connecting WhatsApp to the Claude Code CLI.

That was OpenClaw v1. One hour of work.

He started using it immediately. Spotty internet in Morocco, but it worked. Translating menus at restaurants. Understanding signs in Arabic. Finding places to eat. A personal AI assistant living in his WhatsApp.

But the revealing moment came by accident. Peter sent a voice message by mistake. Instead of failing, the agent recognized the file type and built voice support on its own. Without anyone programming it. The AI agent detected it was receiving audio and wrote the code to process it.

If that doesn't give you chills, you're not paying attention to what's coming.

The naming disaster (because there's always one)

Peter already said it: he's bad with names. PSPDFKit was proof enough. But with OpenClaw, the universe decided to turn it into comedy.

First name: Clawdbot. It worked. It made sense. Until Anthropic sent a trademark complaint — too similar to "Claude."

Second name: MoltBot. OK, less creative, but viable. Until he registered the domain and within seconds — literally seconds — crypto scammers hijacked related domains.

Peter was about to delete everything. Cancel the project. Forget about it.

But he didn't. He did a secret rebrand: OpenClaw. And this time it stuck.

The moral: if your project survives three naming attempts, a trademark complaint, and a crypto scammer attack, it was probably meant to exist.

From builder to leader at OpenAI

Today Peter Steinberger leads the personal agents division at OpenAI. His mission is clear: "Build an agent that even my mom can use."

"I can perfectly see how OpenClaw could become a huge company. And no, that doesn't really excite me. I'm a builder at heart." — Peter Steinberger

Revealing detail: Peter has been a heavy user of OpenAI Codex since May 2025. He built tools like CodexBar to monitor his own usage. And even though he created OpenClaw on top of Claude Code, he prefers Codex for programming. He has no tribal loyalties — he uses whatever works best for each task.

Lessons for builders

Peter Steinberger's story isn't a playbook you can copy. It's something better: it's evidence of patterns that repeat.

1. Bootstrapping works. It's not glamorous. It doesn't make TechCrunch. But ten years of organic growth generated a €100M+ exit without diluting a cent.

2. Burnout comes after success, not just before. Prepare for the day after you win. Very few people do.

3. Quantity precedes quality. 43 failed projects before OpenClaw. You don't find your best idea by optimizing. You find it by building a lot.

4. The best products are born from real needs. OpenClaw wasn't born from a market analysis. It was born from needing to translate a menu in Marrakesh with bad internet.

5. The path is never linear. Student → teacher → engineer in SF → founder → CEO → burnout → traveler → experimenter → creator → leader at OpenAI. If your path looks messy, you're probably doing fine.

6. Build in public, even when it's uncomfortable. The ridiculous names, the scammers, the Anthropic complaint — it all became part of the story. And the story is what makes people connect with your product.

---

Peter Steinberger didn't plan to create OpenClaw. He didn't plan to burn out. He didn't plan to end up at OpenAI. But he never stopped building — even when building meant 43 projects nobody was going to use.

If you're on project 5, or 15, or 40 — keep building. Number 44 might be the one that changes everything.

Related content

• 📘 AI Agents in 2026: Complete Guide — Understand the complete agent ecosystem that OpenClaw helped create
• 📘 How to Set Up Your First AI Agent with OpenClaw — Try OpenClaw yourself: step-by-step tutorial
• 📘 Agentic Engineering: The Complete Framework — The professional methodology born from tools like OpenClaw