Notes on building AI prototypes, from real projects.
Most AI ideas don't fail because the technology doesn't work — they fail because nobody found out cheaply whether they were worth building.