{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreidnawabhd6dt4voycmqlwcgcxqukt6zufgsjrwkuuz67magufcplm",
"uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mnmph7l5xia2"
},
"path": "/t/codex-has-recently-started-ignoring-instructions-and-making-unsafe-patches/1382793#post_5",
"publishedAt": "2026-06-06T12:31:20.000Z",
"site": "https://community.openai.com",
"textContent": "Small update:\n\nI switched to GPT-5.5 in Codex and, so far, its behavior is back to normal.\n\nWith the same project and the same workflow, GPT-5.5 follows instructions much better than GPT-5.4 did during the last few days.\n\nI am now using Codex with a very strict workflow:\n\nbackup → one change → list files → approval → patch\n\nThis seems to work well again with GPT-5.5.\n\nOf course I cannot know what exactly happened with GPT-5.4. It may have been a regression, a routing issue, or some change in model behavior.\n\nHowever, from a user’s point of view, the experience is quite clear: GPT-5.4 suddenly became much harder to control, while GPT-5.5 works properly again. This inevitably creates the impression that users are being pushed, indirectly, toward the newer and more expensive model.\n\nI am not stating this as a fact, only as an observation from real usage.\n\nΣτις Σάβ 6 Ιουν 2026 στις 1:16 π.μ., ο/η Paul <notifications@openai1.discoursemail.com> έγραψε:",
"title": "Codex has recently started ignoring instructions and making unsafe patches"
}