Claude Code quality regressions and fixes
Key Questions
What quality issues affected Claude Code in March and April?
Anthropic addressed problems with verbosity and reasoning that users reported during that period. The company has since focused on ongoing stability improvements.
How are users tuning Claude Code for better production results?
New tuning approaches include Karpathy's CLAUDE.md rules along with skills and subagent fixes. These help refine outputs for more reliable production use.
What do the related articles reveal about Claude Code versus Codex?
Hands-on comparisons show differences in models, pricing, speed, and token costs after weeks of testing. Anthropic has publicly acknowledged earlier code quality regressions raised by clients.
Anthropic addressed verbosity/reasoning from March-April; ongoing stability. New tuning via Karpathy CLAUDE.md rules, skills, and subagent fixes for production.
Sources (2)
Updated May 22, 2026