Each lesson teaches one idea, grounded in the wiki's corpus of 600-plus annotated talks and every claim traceable to its source. Some stand alone. Others are stops on a guided path — a sequence meant to be walked in order, each lesson building on the one before. A path is the teaching counterpart to a theme: where a theme is a reference map you consult, a path is the tour you are led through.
The Road to AGI
A five-lesson arc on the central argument in AI: does scaling — bigger models, more data, more compute — carry all the way to general intelligence, or is something missing that scale cannot supply? Start at the beginning; each lesson assumes the last. Its reference companion is the Road to AGI theme.
- The Bitter Lesson Why general methods that ride raw compute keep beating clever, hand-built knowledge.
- Scaling Laws The empirical rule that turned 'bigger is better' from a hunch into a plan you can bank on.
- Verifiability Why AI surges wherever an answer can be automatically checked — and stays jagged where it can't.
- The Generalisation Gap The strongest objection: why a model can ace a hard exam yet fail what a five-year-old does easily.
- Is Scaling Enough? The resolution — separating the settled question (does scaling work) from the open one (does it suffice).