The cloud system logged more than 1,700 automated investigations and root caused over 600 cases in three months.
OpenAI built an autonomous system called the Performance Factory that finds and fixes performance regressions in ChatGPT. The system runs in the cloud, where Codex agents investigate slowdowns, record findings in a task, propose fixes, and verify results after deployment.
Engineers define recurring workflows and write instructions for the Codex agents that carry them out. Before Codex, engineers compared traces by hand, debugged live machines, and read code, and the team built more than 30 performance tools. The Regression Workflow tracked more than 1,700 investigations and root caused over 600 cases in roughly three months, or 17 automated investigations a day.
Service teams that adopt agentic workflows gain a repeatable pattern: an agent investigates an alert, records evidence, drafts a fix, and then checks the deployed change against baseline. The approach matters for operators who cannot staff manual triage at agentic development speed, and it lowers the expertise needed to find where to look in an unfamiliar system.
OpenAI calls the Regression Workflow its first, and the company says it will explain how its Codex usage grew into the Performance Factory and share early lessons from running agent workflows in production. OpenAI describes the factory as a way to keep pace with its roadmap, so watch how far the workflow model extends beyond regression work.
What matters
- OpenAI’s Performance Factory automates regression investigations, fixes, and post-deploy checks.
- Teams adopting agentic workflows can copy the pattern of agents that investigate and propose fixes.
- The next signal is how far OpenAI extends the factory beyond regression work to other recurring jobs.
Why it matters
The next signal is how far OpenAI extends the factory beyond regression work to other recurring jobs.
This GenAI News article was prepared in original wording using reporting and materials published by OpenAI News. Source reference: https://openai.com/index/how-openai-performance-factory-keeps-chatgpt-fast.
Drafted by the GenAI News review pipeline.
