Lobster Daddy Burns $1.3M in Tokens Monthly! Wouldn't Afford It Without Joining OpenAI
Peter Steinberger, known as 'Lobster Daddy', revealed his CodexBar dashboard showing $1.3 million (about 9.4 million RMB) in OpenAI API costs over the past 30 days, consuming 603 billion tokens. Most of it goes to developing OpenClaw with a team of 3 running 100 Codex agents in collaboration. He claims fast mode can be turned off to cut costs by 70%, but still considers it cheaper than hiring engineers.
Peter Steinberger, widely known as "Lobster Daddy" in the AI community, recently posted a screenshot of his CodexBar dashboard that stunned the internet. Over the past 30 days, his OpenAI API usage racked up a staggering $1,305,088 (approximately ¥9.4 million), consuming 603 billion tokens across 7.6 million requests. The most frequently used model was GPT-5.5. The best part? All these costs were covered by OpenAI itself, as Steinberger is now an employee there.
Most of these tokens went into developing OpenClaw, a project where Steinberger's team — just three people — runs about 100 Codex agents simultaneously. These agents collaborate autonomously: they review PRs, scan for security vulnerabilities, detect duplicate issues, write fixes, monitor benchmark regressions, and even listen to meetings to create PRs. The agents also review each other's work, forming a self-policing multi-agent system. Steinberger is candid that some of the messier PRs he fixes were likely created by Claude (from Anthropic), which the team also uses alongside tools like Clawpatch.ai, Vercel's Deepsec, and Codex Security.
The dashboard revealed about 206,000 requests per day, averaging 2.4 calls per second. This indicates the agents are running continuously and stably. Steinberger explained that the high token consumption is due to "fast mode" designed for rapid, high-frequency collaboration. If he turns it off, costs drop by 70% — but that still leaves a $400,000 monthly bill. Yet he argues this is not expensive when compared to hiring a team of senior engineers, especially in San Francisco where a single senior engineer can cost hundreds of thousands annually. He emphasizes that OpenClaw is open-source and compatible with mainstream models, providing a high return on investment.
The trend is clear: software development is shifting from humans using tools to humans managing AI agent teams. Steinberger's approach echoes Sam Altman's vision of an "extremely multi-agent" world. And when asked about the remaining tokens not used for OpenClaw, Steinberger hinted that he's also working on several new startup projects, leaving the community eagerly anticipating what's next.