Teach yourself anything
What this is
Chapter 9 makes a point that most people miss about AI: it can be the kind of coach you’ve probably never had. Not because it’s smarter than a human coach, but because it remembers what you said you wanted, notices when you’re drifting from it, and is available every time you actually have ten minutes — not just when you’ve booked a session.
You saw Rosa do this in the book. Six years of landscape photography, then a regional competition judge wrote that her compositions were “safe” — nothing stood apart. She opened AI and described the plateau. AI’s first move was to ask to see her work. The feedback — three of her five photos placed the subject at exactly the same point in the frame; her horizons were always level and always at the same height — wasn’t something she could have noticed from inside her own habits. Over the next few weeks, AI set specific challenges and pointed out what she changed and what she reverted to without noticing. Three weeks in, the photo she almost didn’t share was the strongest image she’d produced. The “safe” compositions had been the ceiling, not the floor.
This exercise starts a coaching relationship like Rosa’s. By the end of the next twenty to thirty minutes you’ll have AI’s first specific observations on something you’re trying to get better at, a concrete exercise to do this week, and an agreed plan for what to send back.
What you’ll need
- 20–30 minutes for this first session
- A skill you’ve been trying to improve at — ideally something you’ve plateaued on, or something you keep meaning to make progress on but don’t
- Your usual AI tool open in a fresh conversation
- If your skill has shareable output (writing, photos, recordings, drafts), have one or two recent examples ready to paste or upload
Pick the right skill
Three filters to make sure you’re picking something this exercise can actually move:
- A skill you’ve been at for a while. Long enough that you have a sense of your usual output and your usual blind spots. If you’re brand new to something, this exercise is less useful — you don’t have patterns yet for AI to spot
- External feedback would help. You suspect you’re stuck in habits you can’t see from inside. Rosa had a judge’s feedback that gave her direction; you don’t need that, but you do need to be open to feedback that’s specific and occasionally uncomfortable
- Improvable through small, repeated work. Examples: photography, structured decision-making, learning a language. Not skills that genuinely require institutional training or qualifications
If your skill produces visible output (an essay, a photo, a recording, a chart, a plan), bring examples to the first session. AI’s most useful feedback comes from looking at the actual work, not at your description of it.
Step 1 — describe where you actually are, honestly
This is the move Rosa made. Not “teach me photography”. An honest description of where she’d been, what she’d been getting, and what “better” would look like.
Be more uncomfortable than you’d like to be in this prompt. If you tell AI a polished version of your skill level, it’ll give you a polished version of advice.
If your skill has shareable output, attach one or two recent examples in the same message as the prompt below. If your work isn’t easily shareable, make sure your description includes a specific recent example.
I want to get better at [skill]. Three things to tell you.
Where I actually am: [what you can do, what you struggle with, and any external feedback you’ve had — even informal, even feedback you didn’t like] How long I’ve been at this level: [time] What “better” would actually look like: [be specific — not “be good at X”, but “be able to do Y in situation Z”]
Don’t give me a reading list or a course recommendation. Don’t tell me to “practise more” — that’s not advice, it’s encouragement. Look at what I’ve shared and tell me what patterns, habits, or blind spots you see.
Read AI’s response carefully. The good version of this response will tell you something you didn’t already know — usually something specific about a pattern across your work that you couldn’t see from inside it. Rosa’s three-of-five-same-frame-position observation is the kind of thing to look for. If AI just gives you back a polished summary of what you said, push back: “That’s a summary of my situation, not a pattern observation. Look at the specifics again.”
Step 2 — start the coaching relationship
Once AI has given you specific feedback, this second prompt is the move that turns a one-off conversation into an ongoing one.
Based on what you’ve just told me, two questions:
What’s the first thing I should work on this week? What should I send you next week so you can tell me whether I’ve improved?
Give me one concrete exercise — not a topic to study, an actual thing to do — that I can complete in the next seven days. Make it specific. Tell me what to pay attention to as I do it.
The two questions are the key — and it’s the framing the book is explicit about. “You’re not asking for a lesson. You’re starting a coaching relationship.”
Step 3 — do the exercise (this week)
Actually do it. Not just read AI’s response.
This is the part of the loop where the value of the relationship gets built. If you skip this — if you bookmark a great-sounding exercise and never act on it — you have a saved conversation, not a coaching relationship. The exercise is the practice. The practice is what the coach watches for.
Step 4 — return with results (in seven days)
When you’ve done the exercise — or as much of it as the next seven days allowed — return to the same AI conversation and bring the results back.
Here’s what I did this week: [describe or paste your work] What I found difficult: [the specific moment something didn’t work] What I think went well: [your honest assessment — not what you’d say to encourage yourself]
What’s the one thing I should focus on fixing before next time? And what’s my exercise for next week?
Two practical tips for the ongoing loop.
Keep it in the same conversation if you can. AI remembers across the conversation — the patterns it spotted in Week 1 are what let it tell you in Week 4 whether you’ve genuinely changed or just briefly performed better. If you have to start a new conversation, paste a short summary at the top: “Last week I worked on [X]. The feedback was [Y]. I’m focusing on [Z] this week.”
Don’t soften your reports. Rosa’s breakthrough came from sharing a photo she thought was a failure. The coaching pattern only works if AI sees the actual results, including the disappointing ones. If you only share what you’re pleased with, you’re back to encouragement, not coaching.
What you’ll notice
AI’s first specific observations should tell you something you didn’t already know. If the feedback sounds generic, your Step 1 description might not have been uncomfortable enough. Try again with more raw detail about what you actually struggle with.
The exercise AI sets will sound smaller than you expected. “Shoot the same scene with the horizon in the top third, then the bottom fifth” isn’t a course. It’s a single, specific, achievable thing. That’s the right scale. If you’re expecting a syllabus, you’re looking for a teacher; if you can accept a single targeted move, you’re working with a coach.
The value isn’t in this session. It’s in returning. Week after week, the same conversation, the same goal, with AI remembering what you’ve been working on and pushing you where you need it. Rosa didn’t become a professional photographer. She got a sustainable way to keep improving — “sustained, specific feedback on her own work from something that remembers what she’s been practising and what she keeps falling back to.”