• The host frames AI as a direct challenge to online educators because learners may ask bots instead of buying courses. 23:22
  • Matt separates teaching into "what" knowledge, such as syntax, and "why" wisdom, such as judgment about when and how to use it. 23:57
  • He argues that knowledge has become cheap to acquire, but wisdom has not become easier to learn. 24:24
  • Total TypeScript revenue went down because people became less interested in that kind of tactical material. 25:02
  • Matt first tried making material about putting AI into applications, but later concluded it was the wrong bet for him. 25:31
  • Around the winter break after December, he saw that agents had become good enough to delegate real work to. 25:58
  • His view is that AI has largely eaten tactical programming, leaving humans responsible for strategic programming. 26:22
  • Experience building agent applications helped him think about data flow, state, prompt placement, compaction, and how agents pass information between steps. 28:24
  • He began treating agent workflows as processes, loops, or finite state machines that move information through multiple steps. 29:08
  • Skills became his preferred distribution mechanism for sharing those workflows. 30:00