Users report AI model quality collapse: skipped steps, blind edits and doom loops
JayJeds · reddit · 2026-10-11
A user reports that over the past week, models have been making careless mistakes on complex tasks: told to make surgical patches, the model jumps from step 1 to step 6 guessing the in-between, then admits the error, revises, and doom-loops on the same code. The author claims OpenAI's model performance fell off a cliff after Anthropic's Opus 5.5 release, with even the top-tier 'Astra' making sloppy mistakes.
More from Models
- Leaks claim OpenAI still has two pretrain runs dwarfing Google's astra — iruletheworldmo · 2026-10-11
- 75-test coding benchmark pits RTX 4090 vs Strix Halo vs M5 Max for local LLMs — julianharris · 2026-10-11
- Anthropic reportedly asked AWS to urgently pull Sonnet 3.5, 3.6 and 3.7 from Bedrock — repligate · 2026-10-11
- Benchmark audits find ~30% of SWE-Bench Pro tasks broken; SciCode and Terminal-Bench flawed too — geoffwolfe · 2026-10-11
- Gemini's Erotica Ban Frustrates Creative Writers: "Does Google Not Want Our Money?" — Original_Being2545 · 2026-10-11
- ChatGPT Pro User Paying $100/Month Hit With "Capabilities Reduced" and Upgrade Prompts — harmaniexchange · 2026-10-11