{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreidsqwf7ttykxcqcbpksf6slt2tbdevpk4ynrx2hdupdrhvpmlfpp4",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mnbeedxj6xg2"
  },
  "path": "/t/perceived-drop-in-gpt-5-quality-over-the-last-few-weeks/1382225?page=2#post_26",
  "publishedAt": "2026-06-01T23:56:32.000Z",
  "site": "https://community.openai.com",
  "textContent": "Two perspectives I can add here:\n\nFirst, after a very good few months with Codex, I recently had a bad stretch where many of my Codex outputs were questionable, to say the least.\n\nIn my case, the reason was me. I had made small, untested changes across my harnesses and degraded performance without noticing it right away. I had to take a step back for a few days before picking up the pieces and learning from my mistakes.\n\nThe second perspective is based on what I have seen here in the forum. There are almost no users agreeing with that sentiment. We have seen literal storms of complaints about model degradation before, but that is not happening here.\n\nThe good news may be that this is one of those learning opportunities based on tough experience. It is the kind of thing people usually do not like to talk about. After all, everyone is seemingly building faster and better all the time. Who has time to publicly reflect when things go the other way?\n\nThis is just me sharing my experience. I hope this helps!",
  "title": "Perceived Drop in GPT-5 Quality Over the Last Few Weeks"
}