# Claude Code fixes 13 of 18 bugs when given evidence, not just error messages

> Original title: Everyone's trying to make AI agents smarter.

- Source: X search: AI launch posts (introducing, just launched)
- Published: 2026-10-27T22:59:30.000Z
- Bad Influence: https://badinfluence.news/items/b987yzlfvn02phxaeb49rn339
- Original: https://x.com/rae1101x/status/2115216850711924820

## Summary

A test by an independent developer found Claude Code fixed the root cause in 6 of 18 runs given only an error message, rising to 13 of 18 when given evidence showing where the bad value came from. The author describes the experiment as small and not a universal benchmark, and argues that agents often add guards that make errors disappear rather than fixing the underlying bug.

## Post

Everyone's trying to make AI agents smarter.

My agent went from 6/18 to 13/18 real bug fixes without changing the model.

The missing piece wasn't intelligence. https://x.com/i/article/2114939319656087552