AWS solved NHL playoff clinching with constraint programming and said no LLM belongs here.

The AWS Generative AI Innovation Center built a two-part solver: a 0-day CP-SAT model on Google OR-Tools testing whether any scenario exists where a team misses, plus an n-day lookahead using custom tree search where each layer is a game day and each node an outcome, calling the 0-day solver at every node. Median runtime in minutes for 1-day scenarios, pruning efficiency near 100%, validated against four NHL seasons where output matched published scenarios exactly. Seven cascading tiebreakers requiring exact verification is precisely where a probabilistic model is wrong, and a team inside a generative AI org said so in public.