🔍 Read the full analysis: What OpenAI’s Performance Factory Does For ChatGPT on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
OpenAI says its cloud-based Performance Factory routes performance alerts through Codex agents, engineers and deployment checks. In just over three months, its first workflow tracked more than 1,700 investigations and identified root causes in over 600 cases, according to the company. OpenAI has not published independent results or a measured change in ChatGPT speed attributable to the system.
OpenAI says it has built a cloud-based system called the Performance Factory to move performance alerts through investigation, engineer review and post-deployment checks, including for services such as ChatGPT, as described in the original analysis. The company reports that its first workflow tracked more than 1,700 investigations and identified root causes in over 600 cases in just over three months, but it has not reported a measured improvement in ChatGPT latency or other user-facing performance outcomes, a distinction relevant to assessing OpenAI’s claims.
The first process, called the Regression Workflow, begins when a performance alert fires. A Codex agent investigates the slowdown and records its findings in a task, then proposes a possible fix for an engineer to review. Once a change is deployed, another agent returns to the task and checks whether performance has recovered to baseline, according to OpenAI’s description.
The system combines a task service, configurable workflows, a dispatch loop that assigns ready work to agents, and cloud execution for Codex. It can also resume a task when an event occurs, such as a review comment or completed deployment. Engineers define the task’s states and permitted transitions, and identify steps that require human attention; agents receive written prompts for each stage.
OpenAI says the workflow’s more than 1,700 investigations amount to about 17 automated investigations per day. The company says it traced root causes in over 600 cases during the reporting period. Those figures describe internal activity, not a published measure of fixes shipped, time saved, or changes in service performance. The company has not supplied case-level data or an independent evaluation.
How OpenAI’s Performance Factory keeps ChatGPT fast
OpenAI says its cloud system routes performance alerts through Codex agents, engineers and deployment checks. The reported activity shows internal scale, but not a measured improvement for ChatGPT users.
From alert to recovery check
The Regression Workflow carries a performance task through investigation, human review and follow-up.
A Codex agent examines the slowdown and records findings in a task.
An engineer reviews the agent’s suggested change.
Another agent checks whether performance returns to baseline.
How the work changed
- Engineers inspected metrics, logs and traces.
- They selected from more than 30 performance tools.
- People guided investigations and follow-up.
- Alerts can start cloud-based Codex investigations.
- Task states and findings carry across stages.
- Engineers review proposed fixes and deployment checks.
What the figures tell readers
OpenAI reports substantial internal investigation activity over just over three months.
More alerts can be worked
A consistent workflow may help teams investigate slowdowns across services such as ChatGPT.
Faster service is unproven
The account gives no measured latency or availability change attributable to the system.
What remains unmeasured
- How many investigations led to confirmed fixes.
- How many proposed changes engineers approved.
- How quickly cases recovered after alerts.
- Whether users experienced better latency or availability.
How Alert Handling Could Change
The system targets the work between detecting a slowdown and confirming that a fix worked. OpenAI says engineers previously needed to inspect metrics, logs and traces, choose from more than 30 performance tools, and apply service-specific knowledge. Codex made it easier to begin investigations, the company says, but engineers still had to guide the work, handle proposed fixes and follow up after deployment.
By carrying a task across those steps, the Performance Factory could reduce manual coordination and let teams process a larger volume of alerts through a consistent routine. For users, the potential benefit is indirect: more systematic investigations may help engineers respond to slowdowns in ChatGPT and other services. However, the reported number of investigations does not establish that users saw faster responses, fewer outages or any other measurable improvement.
The distinction matters because operational throughput and service outcomes are different measures. An investigation can identify a cause without leading to an approved change, and a deployed change does not by itself show that recovery was faster. OpenAI’s account describes a system intended to support reliability work, while leaving its effect on customer experience unquantified.
Popular right now

Ninja Foodi DZ201 8-Qt DualZone 6-in-1 Air Fryer

Dyson Cyclone V10 Origin Cordless Vacuum

De'Longhi Classic Signature Espresso Machine with Adjustable Milk Frother
As an affiliate, we earn on qualifying purchases.
From Manual Debugging to Codex
OpenAI describes the Performance Factory as the latest step in a series of changes to its performance work. Before using Codex, engineers manually compared metrics and traces, debugged live machines and read code to locate causes. The company says recurring issues, including regressions in ChatGPT first-token latency, led it to develop more than 30 specialized tools.
Engineers later used Codex to search codebases, locate relevant signals and profile services. OpenAI says it gathered investigation steps in a shared skill called perf-forensics, covering questions such as when a slowdown began, which requests were affected, and whether a deployment or configuration change might explain it. The company says it refined the skill using evaluations based on past regressions.
OpenAI then used cloud-based automated investigations to start Codex sessions from alerts. But people still had to turn findings into fixes and track what happened afterward. The Performance Factory is intended to connect those stages, retaining task state and findings as work passes between agents and engineers.
What the Report Does Not Measure
OpenAI has not specified the calendar dates covered by the just-over-three-month period, provided the underlying alert and case data, or explained precisely how it counts an investigation or a root cause. It also has not said what share of investigations produced a confirmed fix, how many proposed changes engineers approved, or how long cases took from alert to recovery.
The account does not quantify whether the system changed ChatGPT latency or availability for users, nor compare its results with performance work conducted without the factory. The reported investigation volume therefore cannot establish that fixes were delivered faster or that users experienced a measurable improvement. No external assessment is described.
Expansion Beyond Regression Alerts
OpenAI says the Performance Factory is designed to support recurring workflows across its services. It has described a possible chain in which a workflow identifies an optimization opportunity and opens a task in another workflow to test and implement it. That would extend the system beyond investigating regressions, but the company has not announced a release date, public feature or timetable for expansion.
The next useful evidence would include deployment rates, recovery times and service-level results, along with clearer definitions of the investigation and root-cause counts. OpenAI has not said when it will publish such measurements. Until then, the factory’s reported activity shows that the workflow is in use internally, while its effect on ChatGPT users remains unverified.
Key Questions
What is OpenAI’s Performance Factory?
It is an internal, cloud-based system that routes performance alerts through Codex investigations, engineer review and checks after deployment, according to OpenAI.
How does the Regression Workflow operate?
An alert starts an agent investigation. The agent records findings and proposes a fix for an engineer to review; after deployment, another agent checks whether performance has returned to baseline.
How many investigations has OpenAI reported?
OpenAI says the first workflow tracked more than 1,700 investigations and identified root causes in over 600 cases during a period of just over three months. The company did not provide the exact dates or underlying case data.
Has OpenAI shown that the system makes ChatGPT faster?
No quantified change in ChatGPT latency or other user-facing performance is reported. The figures describe investigation activity, not measured customer outcomes.
Will the Performance Factory become a public ChatGPT feature?
OpenAI describes it as an internal system for operational workflows and has not announced a public release, rollout date or timetable for expansion.
Primary source: OpenAI · via ThorstenMeyerAI.com
