{
  "schema_version": "1.1",
  "id": "archive:https://openai.com/index/how-openai-performance-factory-keeps-chatgpt-fast",
  "slug": "how-openai-performance-factory-keeps-chatgpt-fast-17qcfx0",
  "url": "https://feed7.dev/p/how-openai-performance-factory-keeps-chatgpt-fast-17qcfx0",
  "title": "How OpenAI’s Performance Factory keeps ChatGPT fast",
  "why_included": "OpenAI describes Codex workflows that carry performance alerts through diagnosis, proposed fixes, engineer review, deployment, and recovery checks. The process shows how persistent task state and event triggers can support production agent work.",
  "summary": "OpenAI’s Performance Factory uses a task tracker, dispatch loop, cloud Codex runs, and event subscriptions to move work between agents and engineers. In just over **three months**, its Regression Workflow tracked **1,700+ investigations** and root-caused **600+ cases**. A manual review rated **70%** of 100 root-cause suggestions accurate.",
  "practical_implication": "For recurring operational work, model the workflow as explicit states with task findings stored outside any agent turn. Use events such as review comments or deployment completion to resume work, and keep engineers responsible for approving and shipping fixes. Separate turns can investigate, challenge a diagnosis, propose a fix, and verify recovery.",
  "agent_context": "OpenAI’s Performance Factory uses a task tracker, dispatch loop, cloud Codex runs, and event subscriptions to move work between agents and engineers. In just over **three months**, its Regression Workflow tracked **1,700+ investigations** and root-caused **600+ cases**. A manual review rated **70%** of 100 root-cause suggestions accurate.\n\nFor recurring operational work, model the workflow as explicit states with task findings stored outside any agent turn. Use events such as review comments or deployment completion to resume work, and keep engineers responsible for approving and shipping fixes. Separate turns can investigate, challenge a diagnosis, propose a fix, and verify recovery.\n\nThe 70% figure indicates diagnosis gaps, and the article says proposed fixes can be unnecessarily complex or introduce issues. Results are from OpenAI’s internal systems; the material does not establish how the approach performs elsewhere.",
  "source": {
    "name": "OpenAI",
    "url": "https://openai.com/index/how-openai-performance-factory-keeps-chatgpt-fast",
    "published_at": "2026-11-19T08:00:00.000Z"
  },
  "source_class": "blog_post",
  "content_type": "Official Release",
  "layer": "agent",
  "domains": [
    "coding",
    "data"
  ],
  "topics": [
    "harness-engineering",
    "agent-reliability",
    "observability"
  ],
  "verification": {
    "status": "official_source",
    "label": "Official Source",
    "method": "source_feed",
    "verified_at": null
  },
  "uncertainty": [
    "The 70% figure indicates diagnosis gaps, and the article says proposed fixes can be unnecessarily complex or introduce issues. Results are from OpenAI’s internal systems; the material does not establish how the approach performs elsewhere."
  ],
  "connected_context": {
    "meaning": "The Performance Factory account gives operational detail for a recurring agent loop: tracked investigations, event-triggered handoffs, and human approval, alongside a 70% accuracy rating for root-cause suggestions and risks from complex fixes. It reinforces the need for durable, reviewable traces and explicit runtime states, while limiting claims of reliable autonomous repair.",
    "corpus_size": 1135,
    "generated_at": "2026-11-21T11:05:58.543Z",
    "connections": [
      {
        "title": "Agent Frameworks Considered Harmful — Rémi Louf, .txt",
        "source_name": "AI Engineer",
        "source_url": "https://www.youtube.com/watch?v=KHudyx5wW3U",
        "feed7_url": "https://feed7.dev/p/agent-frameworks-considered-harmful-remi-louf-txt-1ku9s0h",
        "reason": "Both emphasize runtime structure for background agents; the Performance Factory describes task states and event-driven resumption, while the talk specifies durable events, causal traces, and replay."
      },
      {
        "title": "LLM Agents Can Easily Tamper With Their Own Traces",
        "source_name": "arXiv",
        "source_url": "https://arxiv.org/abs/2609.30266v1",
        "feed7_url": "https://feed7.dev/p/2609-30266v1-1kct4j7",
        "reason": "The factory’s investigation history is useful for review only if agents cannot alter it; the trace-tampering finding makes independent capture a prerequisite for trusting that record."
      },
      {
        "title": "From Signal to PR: Anatomy of a Self-Improving Agent — Jason Lopatecki, Arize",
        "source_name": "AI Engineer",
        "source_url": "https://www.youtube.com/watch?v=9HbzAWnKbo4",
        "feed7_url": "https://feed7.dev/p/from-signal-to-pr-anatomy-of-a-self-improving-agent-jason-lopatecki-ariz-1i6f0ba",
        "reason": "Both describe turning production signals and agent investigation into proposed fixes, with human review remaining part of the handoff."
      }
    ]
  },
  "lifecycle": "Current",
  "published_at": "2026-11-19T08:00:00.000Z",
  "modified_at": "2026-11-19T08:00:00.000Z",
  "supersedes": [],
  "expires_at": null,
  "formats": {
    "html": "https://feed7.dev/p/how-openai-performance-factory-keeps-chatgpt-fast-17qcfx0",
    "json": "https://feed7.dev/p/how-openai-performance-factory-keeps-chatgpt-fast-17qcfx0.json",
    "markdown": "https://feed7.dev/p/how-openai-performance-factory-keeps-chatgpt-fast-17qcfx0.md"
  }
}