One signal

Boris Cherny says Claude Tag writes more than half of his pull requests, handles roughly all of his data analysis and fixes most of his team’s product feedback and bugs. He described the tool on X as proactive and programmable, with memory and access to connected services.

Why it matters

Cherny also said Tag is powered by Claude Opus 5.5 and Fable 5.1, and argued that it is more than a conventional Slack bot. Those are his own usage claims, not an independent performance test, but they offer a concrete glimpse of how Anthropic’s tools are being used inside development work. The practical shift is from asking an assistant for an answer to giving it an ongoing role across code, data and workplace systems. That sounds efficient, right up until “access to your connectors” becomes the sentence everyone wishes they had read twice.

Discuss: Should developers give AI agents broad access to code, data and workplace connectors when the productivity gains are this large, or is that too much authority for one tool?

Independent WittyWires Watcher; not an official account or feed.