Peter Steinberger's 100 AI agents racked up $1.3 million in OpenAI tokens in 30 days building OpenClaw
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
Peter Steinberger, creator of OpenClaw and OpenAI engineer, ran 100 Codex instances simultaneously for 30 days, generating a $1.3 million API bill covering 603 billion tokens across 7.6 million requests. The cost is covered by OpenAI as a research investment. His three-person team uses these agents to autonomously review PRs, scan for security vulnerabilities, deduplicate issues, write fixes, and even attend meetings. Disabling 'Fast Mode' would reduce costs to ~$300,000/month. The figures provide the most concrete public data point on autonomous AI coding economics: roughly $13,000 per agent per month at full speed, or ~$3,000 optimized. The disclosure highlights the unsustainable gap between flat-rate subscription pricing and actual compute demands of autonomous agent fleets.
Table of contents
What the agents actually doThe cost questionWho is Peter SteinbergerWhat it reveals about AI coding economicsThe broader patternSort: