Claude Code, 2월 업데이트 이후 복잡한 엔지니어링 작업에서 사용 불가 수준으로 품질 저하

Claude Code's February update significantly reduced Extended Thinking token allocation, causing up to 73% degradation in reasoning depth for complex engineering tasks. This quality regression manifested as incorrect behavior patterns—skipping code review, premature completion claims, and shift from careful research to hasty edits—directly impacting power users running large-scale system programming projects. The incident highlights a critical need for transparency in token allocation and a dedicated 'Max Thinking' tier to serve users with complex, multi-step workflows that require deeper reasoning.