Metacogniting

Thinking better, with AI

Every session of this course generates a conversation between two AI hosts, made from that week’s own material — the slides, the transcript, what actually happened in the room. They read it, argue with it, and talk it through in about twenty minutes.

It is the course’s own subject, performed on the course itself. That is the joke and it is also the point: you are listening to a machine metabolise a class about how to think alongside machines.

The hosts are not people. They are generated by Google NotebookLM from session material, and their voices are synthetic. Nobody in the audio is a student, and no student’s words are quoted, paraphrased, or reconstructed — every personal story in an episode is the instructor’s own, told to the room.

We disclose this at the top of every episode and in its metadata, because Apple Podcasts requires it of synthetic-voice shows and because it would be a strange course that hid it. The audio also carries Google’s SynthID watermark, so it is permanently identifiable as machine-generated whether we say so or not. We would rather say so.

It is not a substitute for the session. The hosts have the material; they were not in the room, and a class is mostly the room. Treat an episode the way you would treat a very well-read friend describing a lecture they did not attend — useful, occasionally wrong, and never the thing itself.

Where an episode states something the course has since corrected, the show notes say so.