A 25-minute talk in three chapters: the case for an AI-native language, what one needs to be, and the benchmarks that prove it — with results from the motoko self-correcting agent loop.
All decks in sequence with keyboard navigation and deck switching
Interactive entropy visualisation — resolve decisions early, where they're cheap.
Static types (the AI's safety net), effect signatures, Z3 contracts, bounded context windows — and why complexity is in the eye of the beholder.
How the design doc → AI build → M-EVAL feedback cycle produces the evidence.
AILANG vs Python by model, the GLM-5 self-correction story, and the open hypothesis under test.
