[MODEL] opus 4.6

Preflight Checklist

• [x] I have searched existing issues for similar behavior reports
• [x] This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Claude modified files I didn't ask it to modify

What You Asked Claude to Do

●to plan my rebuilt of the model - to follow my claude.md and think properly... really getting annoying these errors all the time costed me 6 hours work

What Claude Actually Did

● I understand your frustration. You're right — my unverified
claim about the coefficients needing rescaling led to a bad
commit that then needed debugging, burning through credits on
a problem I created.

What happened:

1. I claimed the regression needed rescaling for full fossil

dispatch — without checking

1. The implementation agent trusted that claim and

recalibrated to 27.04/3.11/-0.84

1. We then spent multiple rounds discovering the 20 Mt gap,

the scaler placement bug, and finally the TWh backtest failure
— all downstream of my original wrong assertion

1. A 30-second test would have shown the original coefficients

already work at full fossil scope

I can't report to Anthropic directly, but you can report this
at https://github.com/anthropics/claude-code/issues — that's
the official feedback channel. The core issue: I stated an
assumption as verified fact, violating the research-first
rule, causing cascading rework.

Expected Behavior

verify claims before suggesting random numbers

Files Affected

Permission Mode

Accept Edits was ON (auto-accepting changes)

Can You Reproduce This?

Yes, every time with the same prompt

Steps to Reproduce

_No response_

Claude Model

Sonnet

Relevant Conversation

Impact

Critical - Data loss or corrupted project

Claude Code Version

opus 4.6

Platform

Anthropic API

Additional Context

_No response_