Peter Steinberger Built a $100M Dev Tool, Burned Out, Then Came Back to Code with AI Agents And Never Looked Back
**In an interview on The Pragmatic Engineer Podcast, Peter Steinberger, creator of PSPDFKit and Claudbot, shared how he went from elite iOS engineer to burned-out founder—and came back with a radically different approach to building software. He doesn’t read most of the code he ships. And he’s shipping faster and smarter than ever.**
From Farm Kid to Founding a Framework on a Billion Devices
Peter’s story starts in rural Austria. His first exposure to computers came via a summer guest, and from there, it was full-on obsession.
“Building stuff almost felt like playing a computer game,” Peter said.
Fast forward a few years, and Peter had bootstrapped PSPDFKit, a mobile PDF framework so performant that Apple used it internally. It powered PDF viewing on over a billion devices, and built a loyal base of developers—not through cold email, but by writing world-class blog posts and obsessing over developer experience.
But success came at a cost. Peter burned out.
“I was working most weekends. As a CEO, you're the trash bin. Everything others can’t solve, you have to fix.”
So, in 2020, Peter sold his shares in the company he’d built over 13 years and walked away from tech—for three years.
The Comeback: AI Agents and a Wildly Addictive Personal Assistant
Peter returned to programming in early 2023, cautiously exploring modern web tools and AI.
The turning point? When he fed his messy personal project into Claude Code, Anthropic’s AI coding assistant, and asked it to write specs, then build code from those specs.
“It was my holy f--- mind-blowing moment.”
That spark led to Claudbot—a deeply personalized AI assistant that lives in your terminal, chats with you via WhatsApp or Discord, and controls everything from your smart lights to your GitHub repos. It’s like Her, but you install it with npx claud.
The wildest part? Peter built it by orchestrating 5 to 10 AI agents in parallel, using Codex (OpenAI’s code interpreter) as his “dev team.”
“I Don’t Read Code Anymore. I Weave It.”
Peter’s new workflow sounds like science fiction:
- He prompts agents to build subsystems.
- Each agent cooks (sometimes for 40+ minutes), reading code, planning, and executing.
- He focuses on architecture, taste, and outcome—not syntax.
“Many details are boring. I care about system design. I don't need to read how the button is aligned or which Tailwind class is used.”
He calls PRs “prompt requests.” CI? Mostly local. Reviews? Optional—his agents test themselves.
The Secret to High-Quality AI Code: Close the Loop
Peter’s most powerful insight?
“Code works well with AI because it’s verifiable. You can compile it, run it, test it. That’s the loop. You have to close the loop.”
So he makes every feature testable, often building CLI test runners and synthetic user flows just so agents can validate their own work. That means he can ship fast, safely.
From Bike-Shedding to Trusting the Machine
At PSPDFKit, Peter obsessed over code quality—spacing, naming, architecture.
Now?
“I care more about the outcome. AI made me a better coder because I had to think about how to structure systems to be testable, not just pretty.”
He’s gone from shipping features monthly to merging 600 commits in a single day—and it’s not slop.
What This Means for Wantrepreneurs and Builders
Claudbot isn’t just a neat side project. It’s a glimpse into the future of software creation—one where a single person, with taste and system knowledge, can build what once took teams of 30.
Here’s what Peter’s story reveals:
- You don’t need to know every line of code. You need to know what you’re building and how to verify it works.
- The bottleneck is no longer typing. It’s prompting, validating, and structuring your thinking.
- AI agents don’t replace developers. They replace repetitive plumbing. Your job becomes architect, editor, and prompt engineer.
- Blogging still matters. Sharing hard problems = attracting world-class talent and building reputation.
Peter’s Advice for New Devs in the AI Era
“Be infinitely curious. Build in the open. Everything is just a good question away.”
He’s not looking for line-by-line coders anymore. He’s looking for builders who talk in prompts, think in systems, and ship with taste.
🧠 5 Takeaways for Founders & Devs
- Think like a builder, not a scripter. Focus on the problem, not the plumbing.
- Prompting is the new programming. Learn how to express what you want clearly and conversationally.
- Optimize for agents, not humans. Structure code so AI can navigate, test, and refactor.
- Use AI to write tests and docs. Peter doesn’t write a line himself anymore—but his projects have the best docs he’s ever had.
- Be open. Be playful. Claudbot started as a WhatsApp joke. It’s now a breakout project with 3,000+ stars in a week.
Jensen Huang Didn’t “Bet the Company” — He Simulated the Future Until It Was Inevitable
