Anthropic has acknowledged reports of degraded Claude Code performance, tracing the issues to three separate changes made between March and April: a reasoning effort default change, a bug in session idle-thinking cleanup, and a verbosity-reduction system prompt. All three were reverted by April 20 (version 2.1.116). The Opus 4.7 upgrade has also drawn mixed developer reactions, with some preferring the previous 4.6 behavior. Anthropic reset usage limits for all subscribers and committed to handling such tradeoffs differently going forward. The restricted Mythos model remains unavailable to most users.
Table of contents
Root of the Problem?Syncing Out Older ThinkingAggregate AggressionThe Test-Time-Compute CurveSort: