OpenAI engineers describe using Codex’s Ultrafast mode for incident response, reporting that some security investigations fell from an hour or two to 10–15 minutes.
OpenAI says engineers are using Ultrafast, a high-speed mode in Codex, as a default tool for incident response. In an account published October 26, 2026, staff describe directing Codex to monitor incident discussions, investigate possible causes and answer follow-up questions while response teams work.
The time savings vary by task. AJ Trbojevic, a datacenter security architect at OpenAI, said that assembling and interpreting structured and unstructured information during security investigations previously took an hour or two. He reported that the work now takes 10 to 15 minutes in the cases he described, and can sometimes happen nearly in real time. These are employee-reported experiences, not a controlled evaluation across all incidents.
The workflow draws on more than one information source. Engineers described using Codex to examine logs and search Slack for references to recent changes or similar symptoms. OpenAI staff also said the faster responses help them stay focused instead of pursuing an investigation manually in parallel while waiting for the assistant.
The account describes uses beyond security response. One engineer said most of his Codex interactions now return in under 30 seconds, a pace he finds useful for quick follow-ups. Another said faster assistance made large code refactors feel less costly. OpenAI also describes staff using persistent Codex conversations for daily planning and delegating repetitive browser tasks such as filling out forms.
The examples are part of OpenAI’s series about its internal use of its own technology and APIs. They illustrate how the company says a faster coding assistant fits into specific staff workflows; they do not establish that the same time savings will apply to other teams or organizations.