Editorial illustration for OpenClaw founder runs 100 AI agents for USD 1.3M/month code, review PRs, find bugs
OpenClaw founder runs 100 AI agents for USD 1.3M/month...
OpenClaw founder runs 100 AI agents for USD 1.3M/month code, review PRs, find bugs
Peter Steinberger is spending $1.3 million a month on OpenAI APIs. That buys him 100 AI agents that write code, review pull requests, and hunt bugs. They even lurk in team meetings and open PRs for features discussed moments earlier.
The monthly bill: 603 billion tokens, 7.6 million requests, and a top model called GPT-5.5. OpenAI picks up the tab, or, more accurately, Steinberger does. He doesn’t flinch.
He’s asking a radical question: what would software development look like if token costs didn’t matter? The answer is a $1.3 million experiment, and whether that’s cheap or expensive depends entirely on what you’re trying to build.
Some agents open PRs based on the project's vision; others monitor benchmarks and report regressions in Discord. Agents even listen in on meetings and start PRs for features the team discusses. The team also uses Clawpatch.ai, Vercel's Deepsec, and Codex Security for bug and security analysis.
In 30 days, the OpenAI API bill hit $1.3 million for 603 billion tokens and 7.6 million requests. The top model was GPT-5.5, and OpenAI picks up the tab. Whether that's cheap or expensive depends on your perspective.
Steinberger defended the cost, saying he's exploring how software would be built if token costs didn't matter.
For $1.3 million a month, Steinberger isn’t just buying code. He’s buying a hypothesis. The hypothesis that when the cost of thinking, of generating, testing, iterating, drops to zero, the very shape of software development changes.
We stop asking “Can we afford to try that?” and start asking “What happens if we try everything?” The 100 agents aren’t a workforce. They’re a probe into a future where the bottleneck isn’t compute, but human attention. The real question isn’t whether this is cheap or expensive.
It’s whether we’re ready for a world where the only limit to what software can do is the speed at which we can decide what we want. Steinberger is paying $1.3 million to find out. The rest of us are just watching the bill.
Related Reading
Research & Benchmarks Google's FACTS benchmark shows 70% factuality ceiling across four tests
Research & Benchmarks Databricks finds multi-step agents beat single-turn RAG by 21% to 38% on STaRK
Research & Benchmarks Nvidia's DLSS 4.5 beta adds 6x Multi Frame Generation for RTX 50 GPUs
From Archives New benchmark shows AI video generators look realistic but lack reasoning
From Archives Researchers train AI model achieving near-full performance using 12.5% of experts
Further Reading
- Stop Wasting Money on OpenClaw Agents (Do this instead) - YouTube — YouTube
- He Asked AI To Make Money. It Did. - YouTube — YouTube
- OpenClaw Investment Guide: Morgan Stanley Lists Top AI Agent Platforms — Investing.com
- Nvidia wraps its NemoClaw around OpenClaw for the sake of security — The Register
- Fere AI Raises USD 1.3M to Put a Self-Improving Trading Agent in Everyone’s Hands — The Wire
