A recent misunderstanding among Claude Code users, where many believed switching to a larger model would increase intelligence, has been addressed by Anthropic. The root cause is confusion between two options: Model selection and Effort level. In March, developers complained that Claude Code suddenly "got dumber," failing to read files, skipping tests, and abandoning tasks. Stella Laurenzo, head of AI at AMD, pulled logs from 6852 sessions on GitHub and found that Claude's thinking volume had dropped by 67% compared to before February, stating Claude can no longer be trusted for complex engineering work. On March 4, to reduce latency, Anthropic lowered the default Effort level in Claude Code from high to medium, a change noted in the update log but unnoticed by most users. After backlash, Anthropic restored the default level on April 7 and reset the usage quota for all subscription users. Anthropic's official breakdown states: Model determines the capability, Effort determines the dedication.