Andrew Ambrosino Says OpenAI Cut Memory Use in Long Local Threads
we're giving you some RAM back. local threads could grow memory usage as they became long, even without starting other processes. @btraut declared war on the problem. enjoy in equally important news, @matteing made his handwriting into a font, which i use below
Context
Long agent sessions that slowly eat memory are a real pain for developers who keep coding assistants running for hours, so this fix matters for day to day use even though it is a small update.
Andrew Ambrosino's X bio says he works on Codex and ChatGPT at OpenAI. Codex is OpenAI's coding agent, which can run local threads on a user's computer. He credited @btraut with the fix.
Deep dive
Andrew Ambrosino (@ajambrosino), whose X bio says he works on Codex and ChatGPT at OpenAI, posted on Nov 4 ET that the team is giving users some RAM back.
According to his post, local threads could grow their memory use as they became long, even without starting other processes. He said @btraut declared war on the problem, and told users to enjoy the fix.
The post does not name the product version, say how much memory was saved, or give before and after numbers. Based on his bio and the mention of local threads, it most likely refers to the Codex app, but the post itself does not name it.
In the same post he added, as equally important news, that @matteing turned his handwriting into a font, and he attached an image written in that font.
The post passed 100,000 views by the time of this scan. A day earlier, Ambrosino posted that users can now send him feedback for all the apps, not just OpenAI's, a post that drew close to a million views without explaining what it meant.
What to expect next
Watch the release notes for the Codex app to see exactly which version includes the fix and how much memory it saves. Developers who run long sessions will likely report whether the slowdowns are gone. More performance updates may follow as OpenAI competes with other coding agents on reliability, not only on model quality. OpenAI has not posted an official changelog entry for this fix in the posts found during this scan.