New in Claude Code: your sessions can now message each other.
Instead of having to re-explain yourself in another session, you can now tell Claude to do it. It sends a summary (not your history or files), and the other session picks it up mid-task.
So, I did some research.
The regression is real. But it's not Claude getting dumber. And you can fix that.
Thinking budgets were adjusted. For complex multi-file work, the default medium effort may not be enough.
Three fixes:
1. /effort high (or /effort max on Opus for hard debugging)
2. ~/.claude/settings.json → "showThinkingSummaries": true
3. CLAUDE.md: "Research the codebase before editing. Never change code you haven't read."
GitHub issue #42796 analyzed 17,871 thinking blocks across 6,852 sessions. The pattern: when thinking depth drops, the model shifts from research-first to edit-first.
Claude didn't get worse. The defaults got conservative.