The six stages
Learning AI is a journey, not a download.
Every meaningful skill follows an arc — curiosity, conviction, struggle, absorption, insight, fluency. This is that arc, mapped for the AI era and grounded in learning science — with one practice to try at every stage.
01
Stage one
Spark of Curiosity
It starts with a jaw-drop moment — the first time an AI does something you didn't think was possible. Curiosity is an information gap: the space between what you know and what you suddenly want to know. Don't let the spark pass. Follow it.
In practice: reproduce the jaw-drop within a day — take whatever amazed you and try it on your own real work while the spark is hot.
Grounded in Loewenstein's information-gap theory of curiosity (1994) and Hidi & Renninger's model of triggered interest (2006). Sources →
02
Stage two
Ethical Crossroads
Before you go deep, decide who you want to be with this technology. What will you use it for? Whose work does it touch? Where are your lines? Ethics isn't a module you bolt on at the end — it's the compass you pick up at the start.
In practice: write your three lines — what you'll always use AI for, what you'll never use it for, and what you'll always disclose. Then tell one person.
UNESCO's 2024 AI Competency Framework puts a human-centred mindset and ethics first, before technique — and MIT's AI literacy curricula interleave ethics with every lesson. Sources →
03
Stage three
Challenge Labyrinth
Then it gets hard. Prompts fail, outputs disappoint, the hype doesn't match your screen. This is the stage where most people quit — and it's exactly where the learning happens. Difficulty that slows you down today is what makes the skill stick tomorrow.
In practice: keep a dead-end log — one line per failure: what you tried, what happened. Ten entries in, the pattern shows itself.
Bjork calls these "desirable difficulties" (1994); Kapur's "productive failure" research shows struggling before instruction beats instruction alone. Sources →
04
Stage four
The Rabbit Hole
Somewhere in the struggle, the switch flips. You lose track of time. You're generating your own questions now, chasing threads nobody assigned you. Situational interest has become personal interest — the rabbit hole is where a curiosity becomes yours.
In practice: book the deep block — ninety protected minutes on one problem you actually care about. Reps, not tabs.
This is Csikszentmihalyi's flow state — total absorption when challenge meets skill — and the shift to "well-developed individual interest" in Hidi & Renninger's model. Sources →
05
Stage five
Breakthrough
The aha moment is real — impasse, then sudden restructuring, then certainty. In AI work, the breakthrough is usually the same discovery: you've internalized the jagged frontier. You know, without checking, where AI is brilliant and where it will quietly fail you.
In practice: write the frontier memo — one page on where AI is brilliant for your work and where it quietly fails. Date it; it will change.
Kounios & Beeman mapped the neuroscience of insight (2014); Harvard's "jagged frontier" study documents the AI version — consultants thrive inside the frontier and stumble just past it. Sources →
06
Stage six
Mastery
Mastery isn't a certificate — it's fluency you no longer have to think about, kept sharp by structured practice and feedback. And in the oldest arc we know, the journey ends with a return: the master's last responsibility is to hand the elixir to someone at stage one.
In practice: teach one thing within a week — show one person one real workflow. Teaching is how fluency sticks, and it's how the cycle restarts.
Ericsson's deliberate-practice research (1993) and the Dreyfus skill-acquisition model define the ladder; Campbell's hero's journey supplies the return. Sources →