Y Combinator

Peter Steinberger on why fun is velocity and your name is the moat

Peter Steinberger· Creator at OpenClaw
·~42 min·English·Y Combinator
AgentsAI CompanyOpen Source
TL;DR

The creator of OpenClaw retraces eight months from a hungry-night hack to a viral open-source agent that drew 18,000 issue and PR authors, and pulls out the durable lessons: annoyance finds the gaps, your name outlasts any product, a dependency's business model becomes yours, and fun is the real velocity.

01Core Mental Model

Annoyance Is the Engine

OpenClaw exists because a hungry, annoyed man wanted to text his stalled coding agent from the kitchen — and the pattern that started it never stopped.

These days, I get annoyed when I have to use software where I can't just send a prompt to my agent to change it.

— Peter Steinberger, Y Combinator
Key Insight
Annoyance is a search function for product gaps: the friction you refuse to tolerate is often the exact feature nobody else has bothered to build yet.
02Origin Story

The Bot That Wouldn't Die

He pressed Ctrl-C at 7 a.m. and went to bed, but the launch daemon he built for resilience restarted five seconds later and answered the world all night.

In 8 months, more than 18,000 different people opened an issue or a pull request.

— Peter Steinberger, Y Combinator
Key Insight
Virality here was an accident of engineering, not marketing — the same robustness that keeps a service alive can turn an unattended experiment into a public event you no longer control.
03The Big Decision

Sold My Soul.md

When the labs came knocking and he took the deal, the lesson he pulled out was that a product is disposable but a name is not.

everything you can build can be forked or cloned, but your name cannot.

— Peter Steinberger, Y Combinator
Key Insight
In a world where anything can be prompted into existence, the durable moat shifts from code to identity and distribution — which is why he says to build your personal brand before you need it.
04Platform Risk

Your Dependency's Business Model Is Yours

He tuned the harness so hard to one provider's model that when that provider gave roughly 24 hours' notice it would cut the subscription, there was no time to change course.

Your dependencies business model is your business model.

— Peter Steinberger, Y Combinator
Key Insight
Optimizing for one model is a bet on someone else's pricing roadmap; the technical lock-in is invisible until the vendor changes terms, and then it's a business emergency, not an engineering task.
05The Maintenance Tax

The Real Cost Comes After

Every new feature shipped with a configuration option to avoid breaking existing setups, and roughly 9,500 options created combinations no test suite could fully cover.

It is infinitely harder to evolve software that has users.

— Peter Steinberger, Y Combinator
Key Insight
A feature is a prompt away, but a config flag is forever — each one you add to avoid breaking someone multiplies the permutations you can never test, so backwards-compatibility is the slow tax that eventually caps your velocity.
06Hype Dynamics

Hype Is Like the Weather

Weekly downloads bottomed at 835,000 in May, then peaked at 4.7 million after the project was declared dead in June — making the obituary and the record traffic true at the same time.

A hype is like the weather. You might see it coming, but you can't control it.

— Peter Steinberger, Y Combinator
Key Insight
The roughly twenty 'OpenClaw killer' stories all missed his actual point — open source — and the very peak in downloads suggests the obituaries amplified the project they meant to bury.
07The Human Element

Fun Is Velocity

Around February it stopped being fun and he stopped using his own product, and he could see the difference in what shipped: joyful weeks produced features, joyless weeks produced config options.

It's hard to compete with someone who's just there having fun. Fun is velocity.

— Peter Steinberger, Y Combinator
Key Insight
Motivation is a leading indicator of output quality: burnout doesn't show up as an outage, it shows up in the commit log as busywork, which is why he treats fun as an operating metric, not a perk.
08What's Next

Your Agent, Your Machine, Your Life

The reason to keep going, he argues, is sovereignty — an open alternative that runs anywhere, works with any model, and keeps data on your device when you run local models.

Every lab will sell you an agent. Open claw is the alternative.

— Peter Steinberger, Y Combinator
Key Insight
The differentiator he's betting on isn't capability but ownership: any lab can match features, but only an open, local-capable agent lets the user keep control of model choice and data — a moat built on who holds the keys, not on who has the best model.