OpenAI unveils Performance Factory for ChatGPT
OpenAI revealed Performance Factory, an internal agent orchestration system powered by Codex that autonomously monitors, investigates, and proposes code fixes for ChatGPT latency regressions. Across its initial rollout, the system tracked over 1,700 investigations and contributed dozens of production patches with human engineers in the loop.
Automating code generation was only step one; closing the loop on continuous telemetry diagnostics and production bug fixing is where agentic engineering delivers real ROI. By wiring Codex directly into observability streams, OpenAI is pioneering autonomous maintenance loops that catch regressions before users even notice.
- –Autonomous root-cause analysis: Agents parse distributed traces, logs, and commit histories upon alert triggers, diagnosing bottlenecks without manual engineer intervention.
- –Verified remediation pipeline: Codex drafts targeted pull requests to address regressions, which human engineers review before automated secondary agents verify latency recovery in production.
- –Production-scale impact: In its first three months, the system handled over 1,700 investigations, diagnosed 600+ root causes, and authored or influenced roughly 50 merged production fixes.
- –Paradigm shift for DevOps and SRE: Demonstrates that coding agents are moving beyond interactive IDE autocomplete toward persistent, event-driven infrastructure workers.
DISCOVERED
57m ago
2026-11-19
PUBLISHED
1h ago
2026-11-19
RELEVANCE
AUTHOR
OpenAIDevs