Working with Gen AI · 2026–27
Five 70-minute lessons about explicit rules, counted patterns, and learned prediction — and about the judgement a person still has to supply.
The spine: rules → counts → learned prediction, with fluency is not accuracy carried across all five days.
Every lesson runs the same four parts: a warm-up question, the lesson activity, an in-class reflection and Classroom submission, and a preview of the next lesson. The decks carry the lesson; the two-page handouts are reference sheets, not required worksheets. Decks include teacher notes and need no downloads, accounts, or logins.
The cat problem, Bletchley Park, ELIZA, and a model you train in three minutes.
It doesn't fool you. In 1966 it fooled people who knew how it worked. Why?
Nobody writes a rule about what to say — only a record of what followed what. What comes out?
Which differences can you actually defend from evidence?
Can you put a claim and its evidence where a stranger could check you?
All five lessons, two pages each. Print at 100%, double-sided, no scaling.
Blank six-zone layout to print, plus the two worked examples used in Lesson 5's warm-up.
Each deck is a single self-contained HTML file on a fixed 1920×1080 canvas that scales to any projector. Nothing is stored, transmitted, or logged — no analytics, no accounts, no student data.