Anthropic's new Opus 5.5 playbook: stop saying "think carefully" and hand over whole tasks
udmrzn · x · 2026-09-24
Anthropic, with Addy Osmani, released an official Opus 5.5 usage guide. The model now reasons before every response and decides for itself how much to think, so prompts like "think carefully" or "step by step" only delay responses without clear quality gains — for simple questions, just ask it to answer directly, and adjust depth via Claude Code's effort setting rather than prompts.
The recommended workflow shifts accordingly: hand over an entire task in one message, define "done" (tests passing, legacy code removed) and when the agent should pause and check in. Early testers ran multi-hour coding sessions with little supervision; send follow-ups while it works, since restarting a session is now the costlier move.
More from coding & agent
- Two Codex agents coded, tested and submitted iOS & Android apps to stores in parallel — burkov · 2026-09-24
- You misunderstand AGENTS.md: how coding tools read project instructions — dotey · 2026-09-24
- Forced fresh-worker swap at 60k tokens: 10/12 Terminal-Bench tasks still pass — key_of_door · 2026-09-24
- Solo Dev Open-Sources Tapioca, a Go Terminal Coding Agent That Runs Fully Local — Practical_Witness_95 · 2026-09-24
- Anthropic launches Claude Marketplace with 2,000+ connectors and third-party agents — claudeai · 2026-09-24
- Nyx Terminal launches for Mac: a terminal built for running many coding agents at once — henrymodis · 2026-09-24