Reddit
Anthropic's Slack 'Claude Tag' Gets ~30% Better at Deciding When Not to Respond, Using Channel Context, Memory and Standing Instructions
An August 13 update to Claude's Slack integration lets Claude Tag read context from across a channel plus its memory and standing instructions to decide whether to jump in or stay silent, with Anthropic claiming Claude is "roughly 30% better at determining when, and when not to, proactively respond." It also routes work into existing threads rather than starting new ones, with proactive replies available at no extra cost. The interesting design signal is that the shipped improvement is restraint — the metric being optimized is suppressed responses, not more of them.
↳ Follow the thread