[BUG] seems as if sonnet has been lobotomised

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Since about a week or so ago, sonnet in claude code is terrible. It will not follow instructions consistently. It gets trapped in reasoning loops. It forgets just about everything after compacting. It has gone from being trustworthy as an independent worker to less use that chatgpt copilot. I am struggling to get it to do anything useful even with continual interactive oversight.

What Should Happen?

It should work reliably and independently as it did 2+ weeks ago.

Error Messages/Logs

Steps to Reproduce

Literally any interactions with sonnet.

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.81

Platform

Anthropic API

Operating System

Other Linux

Terminal/Shell

Xterm

Additional Information

_No response_