[Fix/Revert] Ready-to-merge branch prepared to restore model precision and auto-verification

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?

Summary

Following the discussion in #96283 and #96308 regarding the regression introduced around Sep 16 (CoWork-Chat merger, loss of auto-verification, and context window summarization degradation), I have prepared and verified a clean revert branch in my fork.

Since external Pull Requests are restricted to collaborators on this repository, maintainers can inspect or cherry-pick the fix directly from here:

👉 Revert Branch: https://github.com/adri22235/claude-code/tree/revert/restore-model-precision

• Scope: 7 files changed, 5 additions, 451 deletions.
• Goal: Restores model precision, pre-merger auto-verification loops, and prevents educational/factual hallucinations (#96329).

Hope this helps the engineering team resolve the regression quickly!

What Should Happen?

Claude should retain its pre-merger precision, auto-verification capabilities, and stable context window summarization without degrading model reasoning or hallucinating constraints.

Error Messages/Logs

Steps to Reproduce

1. Start a session with specific constraints (e.g., character limits, fact verification, or multi-turn reasoning).
2. Accumulate context to trigger context window summarization or CoWork agent workflows.
3. Observe that the model fails to self-verify constraints before emitting tokens (loss of auto-verification) and exhibits reasoning degradation as reported in #96283 and #96308.

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

latest

Platform

Other

Operating System

Ubuntu/Debian Linux

Terminal/Shell

Other

Additional Information

_No response_