Anthropic has added a built-in Concise output style to Claude Code. Turn it on, and Claude leads with the answer — no opening pleasantries, no recapping its process, no repeated summary. Simple questions default to a few sentences, but the actual task itself doesn't get shortchanged.
The update is aimed at the recent flood of complaints that Claude talks too much. Much of the feedback has centered on Opus 5: a simple question can return a dozen paragraphs, full of long sentences and jargon. Ask for brevity and you'll get it for one turn — then it slides right back. Even putting instructions in CLAUDE.md doesn't reliably rein it in.
Boris Cherny, the creator of Claude Code, said in the comments that Concise is only a temporary patch; the team is still working on a longer-term fix.
That's because Concise is, at heart, a set of system prompts. Claude Code's official docs are explicit that Output Style changes the system prompt, not the model itself. For now, users get a 'talk less' switch. What the actual long-term fix looks like, Anthropic hasn't said yet.
https://twitter.com/ClaudeDevs/status/2090245922685063634