Opus 4.6 Keeps Breaking Down for a Month Straight
Anthropic admits Claude Code got dumber and explains why.
“AI Disruption” Publication 9600 Subscriptions 20% Discount Offer Link.
The AI community is buzzing today. First, OpenAI made a big splash by releasing GPT-5.5, followed by the long-awaited launch of DeepSeek v4.
Right in the middle of all this excitement, Anthropic decided to stir things up too — not with a new model, but with a serious official statement: Yes, you’re right, Claude Code has gotten “dumber,” but there are reasons behind it…
“In the past month, some users reported a decline in Claude Code’s quality. We investigated and have published a post-mortem report summarizing three key issues we identified.”
The story actually goes back to early March. At that time, users on Hacker News, Reddit, and other forums began complaining that Claude Code was becoming slower and increasingly “dumber,” with some bluntly stating that “Claude Code has degraded to the point where it’s no longer usable for complex engineering tasks.”
