Anthropic · Claude · Claude Code · Microsoft · GitHub · GPT · The Register
Anthropic admits it dumbed down Claude when trying to make it smarter
Compiled by KHAO Editorial — aggregated from 1 source. See llms.txt for citation guidance.
◌ Single Source
Claude users who complained about the AI service producing lower-quality responses over the past month weren’t imagining it.
Key facts
- This was fixed April 10 for Sonnet 4.6 and Opus 4.6
- The latest Claude Code build, v2.1.118, defaults to "xhigh" on Sonnet 4.6
- But after shipping the amended system prompt in conjunction with the release of Opus 4.7, subsequent ablation tests, which involve removing system prompt instructions to measure the effect
- First, on March 4, Anthropic adjusted Claude Code's default reasoning effort level from high to medium
Summary
Anthropic on Thursday published the results of a company investigation that found three distinct changes in March and April made things worse for customers using Claude Code, the Claude Agent SDK, and Claude Cowork. Claude's API, the company says, was not affected. Claude users complained bitterly about the quality of Claude's output during March and April, and service availability problems only made matters worse. First, on March 4, Anthropic adjusted Claude Code's default reasoning effort level from high to medium.