Artificial intelligence news.
Straight from the original announcements.

OpenAI builds Performance Factory to automate ChatGPT slowdown fixes

November 19, 2026, 09:00By the AI News Bites Editorial TeamSource: OpenAI

OpenAI has built the Performance Factory, a cloud-based system that uses Codex agents to investigate, fix and verify performance regressions affecting its services. The company says its first workflow tracked more than 1,700 investigations and root caused over 600 cases in just over three months.

Key points
  • OpenAI built the cloud-based Performance Factory to automate performance engineering workflows.
  • Its Regression Workflow investigates alerts, proposes fixes and verifies recovery after deployment.
  • The workflow tracked more than 1,700 investigations and root caused over 600 cases in just over three months.
  • The system combines a task service, dispatch loop, cloud Codex execution and event subscriptions.

From alerts to verified fixes

The Performance Factory automates the process from a performance alert to recovery verification. When an alert fires, its Regression Workflow opens a task. An agent investigates the slowdown, records its findings and proposes a fix for an engineer to review.

After deployment, another agent checks whether performance has returned to baseline. The workflow moves through the stages of needs investigation, root caused, fix proposed and resolved. Engineers can follow tasks in Slack or the workflow’s user interface, add context and review proposed changes.

According to OpenAI, the workflow tracked more than 1,700 investigations and root caused over 600 cases in just over three months. The company describes that as 17 automated investigations every day.

How OpenAI’s process evolved

OpenAI says performance work previously required engineers to examine metrics and logs, compare traces, debug live machines and read code. The team eventually accumulated more than 30 performance tools, including CPU and allocation profilers, to help investigate recurring problems such as regressions in ChatGPT’s first token latency.

Codex initially worked as an engineering partner. It could search large codebases, find relevant metrics and traces, follow interactions across services and perform profiling, but engineers still had to direct the investigation.

The team then captured its process in a shared perf-forensics skill. The instructions tell Codex which tools to use and how to examine when a slowdown began, which requests were affected, and whether deployments or configuration changes could explain it. OpenAI tested and refined the skill with evals based on past regressions.

The next steps moved investigations into the cloud and connected them to broader workflows. Rather than stopping after an agent posted a diagnosis, the Performance Factory tracks what each task needs next and assigns an agent when that step is ready.

Four components run the factory

OpenAI says the system has four main parts: a task tracker with configurable workflows, a dispatch loop, an API that runs Codex in the cloud, and subscriptions that wake agents when events occur.

Engineers define workflows through configuration files and Markdown prompts. The configuration specifies task states, allowed transitions and steps requiring an engineer. The dispatch loop checks which tasks are ready, starts a Codex turn with the instructions for the current state, and records the agent’s progress.

Codex runs in an environment inside OpenAI’s network with access to code repositories and observability tools, including metrics, logs, traces, CPU profiles and memory snapshots. Agents operate under a service identity, allowing OpenAI to scope permissions and identify actions taken by performance agents.

Subscriptions allow tasks to wait for events such as a GitHub review comment, a failed CI check or a completed deployment. When a matching event arrives, the task becomes ready and the dispatch loop can assign an agent to continue the work.

Why it matters for ChatGPT operations

OpenAI’s announcement shows how the company is applying Codex beyond individual coding assistance, using agents to coordinate a repeatable engineering process across investigations, code changes, reviews and production checks.

For OpenAI, the system is intended to help its teams keep pace with what it calls the “staggering pace of agentic development.” For users, the direct relevance is the company’s effort to detect and address service performance problems through an automated workflow, although the announcement does not provide user-facing performance improvements or availability figures.

Related news

GPT-6.2 Astra and Sol Put to the Test Building a 3D PyramidTop story
ModelsNovember 19, 2026, 16:07

GPT-6.2 Astra and Sol Put to the Test Building a 3D Pyramid

Kilo tested OpenAI’s GPT-6.2 Astra and GPT-6.2 Sol by asking both models to build an interactive 3D scene of workers constructing a pyramid. Both completed the project…

GitLab says Duo Agent Platform credits will go up to twice as far
Products and toolsNovember 19, 2026

GitLab says Duo Agent Platform credits will go up to twice as far

GitLab says its Duo Agent Platform now uses fewer model requests for agent and flow tasks, allowing most customers’ GitLab Credits to go up to twice as far. The new…

OpenAI and Databricks bring governed enterprise data to AI tools
BusinessNovember 13, 2026, 16:30

OpenAI and Databricks bring governed enterprise data to AI tools

OpenAI and Databricks are expanding their partnership to help enterprises use ChatGPT, Codex and GPT models with governed business data. The companies describe three…