Workshop OS

The room is the interface.

One stated goal writes the agenda — and every phase rebuilds the room around it.

Why this room exists

Most meeting software gives a group one static surface, whatever the hour is for. Two research threads point the other way. CoExplorer (Microsoft Research, DIS 2024) generated meeting phases from a stated goal and reconfigured the interface for each phase — and found that people want the system to propose structure while they keep the wheel. A CHI 2024 field study of AI-augmented brainwriting placed a language model inside the classic rotating-sheet format: valuable in divergence as one contributor among humans, and valuable in convergence only as a sorting aid that people can visibly overrule.

Workshop OS builds those two findings as one continuous instrument. The goal writes a first agenda. Concerns sent ahead of the session are routed into it by legible rules. Each phase rebuilds the room. The assistant writes cards in the rounds and scores options before the decision — and every human override of its ranking is logged and kept in the record. The wrap-up assembles itself from what happened in the room, in the spirit of ambient-capture tools: the record is a by-product of the work, not a second job.

What is authored, what runs live

Maya and Jonas are scripted colleagues drawing on curated idea pools written for each sample goal. The assistant's cards, scores and announcements are authored ahead of time; the grouping suggestion is a plain word-overlap heuristic. Live in this page: the phase engine, the sheet rotation and timers, the routing rules, the override log, and the wrap-up built from session state. The timebox warning speaks with a real rendered voice, embedded in the page; phase changes are marked with short tones. The seams where authored material sits are exactly where production models would go.

Sources

AI-augmented brainwriting: investigating the use of large language models in group ideation. CHI 2024. doi 10.1145/3613904.3642414
CoExplorer — phase-adaptive, goal-driven meeting environments. DIS 2024, Microsoft Research. doi 10.1145/3643834.3661507
Ambient capture as an industry pattern: notes that assemble themselves while people work.

Fifth study in a five-part series on people and models working together in creative and service work.

RockBloom. Labs Part of the Anthropic Startup Program