The Descent Map
The one-page companion to the video. Everyone learns AI upside down — math, theory, a twelve-week Python course — and quits at week three having built nothing. This map flips it: you don't climb up to AI, you drop straight down into it and learn each layer only at the exact moment you need it. THE ONE RULE (carved at the bottom of the shaft): never learn a thing until something you're building needs it. THE SHAFT — SIX DEPTHS, TOP TO BOTTOM: SURFACE (0m) — pick ONE annoying real problem and solve it end-to-end today with Claude or ChatGPT; that problem is your compass. DEPTH 1 — Drive the tools: three tools not fifty (one thinking model, one coding copilot, one media model) and the three prompting moves that carry almost everything — Context, Constraint, Example. DEPTH 2 — Wire them together: turn prompts into agents; connect a model to your email/calendar/notes and automate ONE daily task so it runs without you (the leverage layer — one prompt saves a minute, one agent saves a morning every morning). DEPTH 3 — Hit a wall, learn why: it forgets → context windows; it lies → models predict, not know; it's slow and costly → tokens. Learn only the one concept that unblocks you, and because you needed it, it sticks for life. DEPTH 4 — Fundamentals, targeted: enough Python to read and change your copilot's code, enough ML intuition to know how models train and break — just the 20% your builds dragged you down to. DEPTH 5 — The bedrock: you ship things other people depend on; every new model is a power tool, not noise; you crossed from user to builder without feeling it. Plus THE 90-DAY VERSION (a real week-by-week plan, not a list of courses) and a fill-in prompting template. Follow the rope. Your descent starts today.
Free. No spam. Unsubscribe anytime.