Claude Code has been updated to run subagents in the background by default, allowing users to delegate multiple tasks concurrently without blocking the main conversation. This change, along with the introduction of Claude Sonnet 5 and Claude Opus 5 models featuring a 1 million token context window, significantly enhances the tool's autonomy for long-running development sessions. A critical security fix was also implemented in August to address a vulnerability in worktree-isolated sessions that could allow destructive git commands against the main checkout, underscoring the need for careful review of autonomous agent outputs. AI
IMPACT Enhances developer productivity by allowing concurrent, unattended task execution and longer context windows.
RANK_REASON Product update for an AI development tool, not a frontier model release.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
