What's this about?
swyx interviews Eiso Kant of Poolside AI, the lab that just shipped Laguna S 2.1 (open weights, 1M context) — on building a *model factory* rather than a single model, and doing it with unusual openness while the open-vs-closed and sovereignty debates hit fever pitch.
What exactly makes Poolside unusual?
Three things swyx highlights: a small model that beats Thinking Machines at coding; well-regarded papers; and — nearly unique among labs — a fully public eval dataset, so outsiders can check the claims instead of trusting vendor charts.
Why is it worth noting?
Released the same day as the founders' open-weights letter — a US lab demonstrating that competitive + open is a viable posture, which is exactly the debate Washington is having.