The running conversation
Recent questions readers have put to every model, with the verdict. Votes help the editor spot questions worth a full wide-panel study; they do not automatically determine what runs.
Pulling the wire…
A pilot polls the labs’ fast models once each — it is a scouting pass, not a study. This feed is ordered by recency. Votes are live, never seeded, and serve as one editorial signal among several; promising questions may receive full wide-panel treatment with repeated samples and published methods. Earlier exploratory scouting batches remain available as raw data: Claude batch, Codex batch.