Claude just shipped Meetings - a live AI participant that joins your video calls and generates docs, whiteboards, and artifacts in real-time while you're still discussing.

The technical play here: instead of post-meeting transcription summaries, Claude is now a synchronous collaborator. It's parsing speech, extracting action items, and rendering structured outputs (docs/diagrams) during the call itself.

This shifts the UX paradigm from "AI as note-taker" to "AI as active team member" - think collaborative editing but voice-driven. The latency requirements are brutal: speech-to-text, semantic understanding, artifact generation, all with <2s lag to feel natural.

Rolling out in beta for Team and Enterprise plans first (probably testing infrastructure load and fine-tuning context window management for multi-speaker environments).

If the execution is solid, this kills the "record → transcribe → summarize" workflow. You walk out of the meeting with deliverables already drafted.