# Claude Code "cannot be trusted" on complex tasks per thousands of coding sessions.

*genai · news · 2026-04-07 · TechRadar*

## Key points

- AMD's Stella Laurenzo claims Claude Code's performance declined after a February 2026 update.
- AMD analysis found stop-hook violations in Claude Code rose from zero to about 10 daily post-update.
- A new 'redact-thinking-2026-02-12' setting correlates with reduced performance on complex engineering tasks.
- Anthropic clarified that Teams and Enterprise users may access higher effort settings for better reasoning.

**Companies:** AMD, Anthropic

[Read the full story on TechRadar](https://www.techradar.com/pro/claude-cannot-be-trusted-to-perform-complex-engineering-tasks-amd-ai-head-slams-anthropics-coding-tool-after-months-of-frustration)

---

Canonical: https://newsio.io/n/6becc7c6-6ecb-40f7-b2ad-73b7ced395a1/claude-code-cannot-be-trusted-on-complex-tasks-per-thousands-of-coding-sessions
Summarized by Newsio from TechRadar. https://newsio.io/how-it-works