Claude Code rebuilds Projects for parallel multi-agent coding; Claude now leads 26% of Anthropic R&D
量子位 · wechat · 2026-09-18
Anthropic redesigned Claude Code Projects into a coordinator-driven multi-agent coding hub: a Coordinator breaks high-level goals into subtasks run by isolated cloud Claude Code instances on separate Git branches that auto-open PRs, with shared memory syncing decisions across threads; beta rolls out to some Pro/Max users. Alongside, Anthropic's new report says 26% of its internal R&D tasks now hit AL4 (AI end-to-end under loose oversight), up from <1% six months ago, with 30,000 concurrent agents, 1B+ monthly decisions (0.002% blocked by oversight), and 6% of R&D compute spent on safety.
More from coding & agent
- Dev builds 3D universe explorer with 20 galaxies, 767 planets using Codex GPT-6.1 Sol — OpenAIDevs · 2026-11-02
- ChatGPT, Claude and Perplexity's fetchers don't run JavaScript, leaving client-rendered sites invisible to AI — gaganghotra_ · 2026-11-02
- SGLang boosts multimodal inference >10% by replacing GPU memory bookkeeping with a dict lookup — charles_irl · 2026-11-02
- kivi: hold Fn, talk, and it types polished text anywhere on your Mac — itsOmSarraf_ · 2026-11-02
- A base LLM plus 11 lines of code is all you need for a decision model — helloiamleonie · 2026-11-02
- Fix codebase structure before asking AI agents to write tests, says engineer — arpit_bhayani · 2026-11-02