AI & agents
OpenAI details its agent-run Performance Factory
- Published
- Source
- OpenAI blog
Summary
OpenAI has shared how it built the Performance Factory, an autonomous system in which Codex agents investigate performance regressions, propose fixes, and verify recovery, while engineers focus on review and deployment. Over roughly three months the regression workflow tracked more than 1,700 investigations and root caused over 600 cases, about 17 automated investigations a day, all per the company's own figures. A separate adversarial Codex session tries to disprove each diagnosis to catch wrong conclusions.
Why it matters for our work
It shows repetitive performance engineering being run by an agent factory inside a major live service, with humans keeping the judgment and approval steps. The post is a concrete example of how human-agent division of labor can speed up large-scale operations.
Translated from the Korean original. Summaries may be translated and edited. Commentary reflects our perspective; forecasts remain the source’s views.