For $1.3 million a month, OpenClaw founder Peter Steinberger runs 100 AI agents that code, review PRs, and find bugs

The Decoder

Original Source

The Decoder

// Analysis

OpenClaw founder Peter Steinberger's team, operating under the wing of OpenAI, is running approximately 100 AI agents to manage and advance software development for the open-source project. This deployment of Codex instances carries a substantial operational cost, reaching $1.3 million in OpenAI API bills over 30 days, attributed to 603 billion tokens and 7.6 million requests, primarily utilizing the GPT-5.5 model. OpenAI currently covers this expense, allowing Steinberger to explore software development paradigms where token costs are not the primary constraint.

// Why It Matters

Signal Intelligence Assessment

This initiative demonstrates the practical feasibility of deploying 100+ AI agents across a comprehensive range of software development tasks, from code generation and PR review to proactive bug finding and feature initiation, setting a new bar for integrated agentic workflows in open-source projects like OpenClaw.
The reported $1.3 million monthly OpenAI API bill for 603 billion tokens provides a concrete, real-world operational benchmark for the cost of advanced, large-scale agentic AI deployment, giving practitioners crucial data for financial planning of similar systems.
The integration of AI agents that monitor benchmarks, report regressions, and proactively initiate PRs based on meeting discussions signifies a tangible advancement in context-aware and proactive agentic contributions beyond reactive coding assistance.
The fact that OpenAI covers the $1.3 million cost highlights the substantial financial barrier to running sophisticated, high-volume AI agent systems, underscoring that current token costs for models like GPT-5.5 remain a critical consideration for wider, independently funded adoption.

Analyzed 17:00 PT · Trace No. 004 · May 2026

Back to Trace No. 004 · May 2026