OpenAI forum report: ChatGPT forgets context then confidently invents it
ombre-ai · reddit · 2026-11-22
A long Nov 19 post on OpenAI's developer forum describes Codex and ChatGPT losing established context — and, crucially, filling the gap: one user's own lyrics were confidently attributed to the wrong song.
- The reporter would rather have a model that "admits uncertainty and preserves task continuity" than one that sounds fluent while losing the thread.
- The post also describes settled design decisions being reopened, with effort going to small details while bigger agreed requirements stay unfinished.
- It's one user's account, not a benchmark, with no data on how widespread it is — but inventing is harder to spot than forgetting.
More from Models
- Inkling 2 mini's 1M-context KV cache dissected: only 8 of 48 layer visits keep full cache — teortaxesTex · 2026-11-22
- GPT-6.2 is hard to steer via system prompt — OpenAI says use it directly in Codex — pvncher · 2026-11-22
- Apple releases Depth Pro 2: 872M metric monocular depth model hitting 95 fps at VGA on B200 — Promptmethus · 2026-11-22
- Community split: 6.2 Astra starts life hated while 5.6 Sol ended up well loved — UltraRareAF · 2026-11-22
- Musk amplifies FSD V8 vs V15 comparison: bounding boxes evolve into photorealistic Reality View — elonmusk · 2026-11-22
- Amazon blocking OpenAI crawlers drives 141% surge in ChatGPT links to Amazon search — gaganghotra_ · 2026-11-22