奇创岛

From «Peak: Secrets from the New Science of Expertise»

Build a Deliberate Practice AI Feedback Coach Toolkit

You'll pick a skill you're actively practicing, then design a reusable AI prompt toolkit using Ericsson's 'purposeful practice' framework — including a performance-diagnosis prompt, an improvement-task generator prompt, a feedback-collection prompt, and a mental-representation calibration prompt — so that AI can serve as your deliberate practice coach at any time.

Final work

A complete '[Skill Name] Deliberate Practice AI Coach Toolkit' (containing 4–6 prompt templates)

Estimated time

60–90 min

Submitted

…

Your final work

Purpose:Internalize Ericsson's 'immediate, specific, weakness-targeted feedback' as reusable AI prompts, so that even without a professional coach you can get high-quality practice diagnostics and continuously recalibrate your mental representations after each daily or weekly session.

Parts:

  • Skill and practice-stage overview (a System Prompt that gives AI the domain context)
  • Work/practice diagnosis prompt (identifies weak spots + evaluates all four purposeful-practice elements one by one)
  • Improvement-task generator prompt (based on the diagnosis, designs specific practice tasks that push you out of your comfort zone)
  • Mental-representation calibration prompt (compares you against a benchmark to locate the cognitive blind spots where you think you got it right but actually didn't)
  • Weekly/daily feedback log prompt (structured summary of each session's gains, problems, and next steps)
  • Plateau-breakthrough prompt (when practice stalls, uses AI to diagnose which layer you're stuck on and provides a breakthrough strategy)

Use cases:

  • · After each practice session, use the work-diagnosis prompt to quickly pinpoint the single biggest issue
  • · Every weekend, use the feedback-log prompt to review the week's practice quality and decide whether to adjust direction next week
  • · Whenever you hit a plateau, use the plateau-breakthrough prompt to re-activate your sense of progress
  • · Keep the toolkit long-term and continuously update the prompts as your skill level rises

Pick a topic

Pick the topic closest to you, or write a custom one when you submit.

Personal Life

Learning / Growth

Family / Parenting

Work / Projects

Tools you'll use from the book

Four-Element Purposeful Practice System Prompt

Embed Ericsson's four elements (clear goal / intense focus / immediate feedback / stepping out of your comfort zone) into AI's System Prompt, so that every piece of AI feedback automatically addresses all four dimensions rather than offering generic advice.

How to use it here:

In your toolkit's 'skill-context System Prompt,' declare: 'You are my deliberate practice coach. Each time I submit work, you must diagnose it against these four dimensions one by one: ① Is my practice goal clear and measurable? ② Did I step out of my comfort zone during the session? ③ Did I receive immediate, specific feedback? ④ Is my mental representation (my internal model of what great work looks like) closer to the benchmark?'

Boundaries:

An overly long System Prompt causes AI to skip key constraints — keep it under 300 words. The four-element diagnostic is a framework, not a rigid equal-weight checklist: AI should prioritize whichever dimension is most relevant to the actual issues in your work.

Mental Representation Comparison Prompt

Mental representation is 'the internalized standard of great work that experts hold.' Use this prompt to force AI to always compare 'your work' against 'the expert standard in this field,' locating cognitive blind spots rather than just listing errors.

How to use it here:

Add the following to your diagnosis prompt: 'Please tell me: in this field, what is the mental representation of a skilled practitioner (their internal model of quality) — and how does my current work differ from it? What are the areas where I think I got it right, but actually deviate from the expert standard?'

Boundaries:

Mental-representation comparison requires AI to have sufficient domain knowledge. For niche skills (e.g., a specific folk instrument or a regional cooking style), AI may provide inaccurate 'expert standards.' In those cases, provide concrete reference examples rather than letting AI define the standard on its own.

Feedback Specificity Prompt

Ericsson emphasizes that feedback must be 'specific, immediate, and targeted at weaknesses.' Use this prompt to constrain AI from giving vague evaluations and require it to pinpoint specific behaviors instead.

How to use it here:

Add a negative constraint to your feedback prompt: 'Do not say things like "overall pretty good," "keep it up," or "there's been improvement" — those are vague evaluations. Your feedback must be precise: ① exactly which specific moment had a problem (e.g., the logic jump in sentence 3 of paragraph 2 — not just "the logic is unclear"); ② what the root cause of that problem is; ③ what the specific improvement task for my next practice session should be.'

Boundaries:

A specificity constraint does not mean 'list 10 problems every time.' More effective deliberate-practice feedback is 'identify the 1–2 most critical issues and improve those,' rather than spreading attention evenly across every flaw.

Plateau Identification Prompt

When practice hits a plateau (you've put in lots of reps but feel stuck), use this dedicated prompt to let AI diagnose 'exactly which layer you're stuck on' — distinguishing whether it's the skill's upper limit, a problem with your practice method, or a mental representation that hasn't been updated.

How to use it here:

Design a 'plateau diagnosis' prompt: 'I've been practicing [skill] for [duration] and lately feel like I've stopped improving. Please help me determine: ① Are my practice tasks still in my comfort zone (naive practice)? ② Is my feedback source not accurate enough? ③ Has my mental representation failed to update (has my standard of "good" not improved)? ④ Or do I need to isolate and drill a more granular sub-skill to break through? Please give me a diagnostic conclusion and a recommended next action.'

Boundaries:

AI's plateau diagnosis is a hypothesis, not a confirmed verdict. You need to run 'an experimental change' based on the suggestion (e.g., try a different feedback method) and observe for 2–4 weeks whether it works — don't treat AI's diagnosis as a final answer and overhaul your entire practice plan immediately.

Coach-Substitute Prompt

Ericsson points out that a great coach is the hardest-to-replicate ingredient of deliberate practice. Use this prompt to have AI simulate 'how the most experienced coach in this field would look at your work,' adding a human-coach perspective.

How to use it here:

Set up the following role in your System Prompt: 'You are now playing the role of a professional coach with 15 years of teaching experience in [skill domain]. You've seen hundreds of learners at every level. Your three specialties are: ① spotting in an instant the places where a learner thinks they've mastered something but has actually developed a bad habit; ② giving direct, specific corrective suggestions without killing motivation; ③ designing the next practice task that is "just slightly too hard" for each learner's current level. Please evaluate my work from this perspective.'

Boundaries:

AI cannot replace the core value of a real coach — a real coach can observe your real-time state under pressure and make immediate adjustments, while AI can only analyze what you explicitly provide. When a skill reaches an advanced stage (e.g., competitive sports, professional music), the marginal benefit of an AI coach-substitute prompt drops significantly; at that point, seek a real professional coach first.

Work rules

Your work MUST include

  • Clearly identify one specific skill you are currently practicing (not vague goals like 'improve myself' or 'learning ability')
  • At least 4 prompt templates that can be copied and pasted directly and impose real constraints on AI (not descriptions like 'ask AI about...')
  • Every prompt must include specific 'negative constraints' (telling AI what NOT to do), not just positive requests
  • The prompts must reflect at least 2 core deliberate-practice concepts (purposeful practice / mental representation / immediate feedback / stepping out of your comfort zone)
  • Explain the usage scenario for each prompt (when to use it) and the expected output format
  • Include a section on 'how to iterate and upgrade the toolkit' (how prompts should evolve as your skill grows)

Your work CANNOT just be

  • Don't just write vague usage suggestions like 'ask AI for writing tips' — every item must be a directly runnable prompt template
  • Don't let AI do the actual practice for you (e.g., 'have AI write the article and then I'll revise it' is outsourcing, not deliberate practice)
  • Don't list book concepts without grounding them in prompt design — there must be a complete transformation chain from book framework to prompt
  • Don't design only one all-purpose prompt — you must create dedicated tools for different stages of the practice cycle (before/during/after/plateau)
  • Don't let AI fabricate practice history or invent false 'you've improved by X' feedback

AI can help you here

Round 1: Help me choose a topic and analyze the toolkit requirements

When to use: You're not yet sure which skill to design the toolkit for, or you're unclear about which type of AI support is most missing in your current practice workflow.

I'm working on the '{{Route name}}' project using *{{Book title}}*. I want to build an AI deliberate-practice coach toolkit, but I'm not sure which skill to design it for, or which type of AI support I most need in my practice workflow.

My situation:
[Fill in: the skill you're practicing, your approximate level, your current practice method, your biggest pain point — is it lack of feedback? Not knowing how to step out of your comfort zone? Or not knowing where your mental representation is wrong?]

Topic options:
[Paste the topic list from the page]

Please output:
1. The most recommended topic, and why it suits my current practice stage
2. Based on my practice pain points, which type of AI support I'm missing most (work diagnosis / mental-representation calibration / making feedback specific / recognizing plateaus)
3. Three key questions I need to think through before designing this toolkit
4. What the final work will look like and how I can use it after it's done

Yellow placeholders need you to fill in before using the AI.

AI can help you organize ideas, but cannot make final judgments for you. Don't let AI fabricate experiences, cases, or misleading content.

Round 2: Help me design the System Prompt and the core diagnosis prompt

When to use: You've chosen your skill and topic, but don't know how to translate the 'purposeful practice framework' into AI prompts with real constraints, or how to write 'negative constraints' that prevent AI from giving vague feedback.

I'm working on the '{{Route name}}' project from *{{Book title}}*.

My topic is: {{Topic}}

The skill I'm practicing and my current situation:
[Fill in: skill name, current level, most obvious practice pain point, the type of problem you most want AI to help you find]

Please help me design these two core prompts for the toolkit:

**Prompt 1: Skill-context System Prompt**
Requirements:
- Establish AI's role as a 'deliberate practice coach' (based on Ericsson's framework)
- Include the four-element diagnostic framework (clear goal / step out of comfort zone / immediate feedback / mental representation)
- Add a negative constraint: 'no vague evaluations allowed'
- Keep it within 200–300 words

**Prompt 2: Work/Practice Diagnosis Prompt**
Requirements:
- Diagnosis must be specific at the behavior level (not 'writing isn't good enough,' but 'the logic jumps in paragraph X')
- Focus on the 1–2 most critical issues (not an averaged list of every flaw)
- Provide 'a specific improvement task for the next practice session' (it must be a task that's just slightly challenging)
- Include a mental-representation comparison (where does my work fall short of expert-level standards?)

Please output both complete, ready-to-use prompt texts, along with a usage-scenario note for each.

Yellow placeholders need you to fill in before using the AI.

AI can help you organize ideas, but cannot make final judgments for you. Don't let AI fabricate experiences, cases, or misleading content.

Round 3: Help me verify the effectiveness of my toolkit

When to use: You've finished a draft of your toolkit and want to check whether the prompt designs truly reflect deliberate-practice principles, and whether each prompt can actually produce useful AI feedback.

I'm submitting my project work for the Shufang Island project.

Book: *{{Book title}}*
Project route: {{Route name}}
My topic: {{Topic}}

My draft deliberate-practice AI coach toolkit:
{{Draft work}}

Please review my toolkit on the following dimensions:

1. **Deliberate-practice principle alignment**: Does every prompt reflect Ericsson's core concepts (the four elements of purposeful practice / mental representation / stepping out of your comfort zone)? Or are they just generic 'give me feedback' questions?

2. **Constraint effectiveness**: Are the negative constraints (telling AI what NOT to do) specific enough? Could AI still give vague feedback after receiving these prompts?

3. **Operability**: Can these prompts be copied and pasted directly, or do they require extensive extra explanation before AI can understand them?

4. **Coverage completeness**: Does the toolkit cover the key stages of the practice loop (at minimum: post-session diagnosis + mental-representation calibration + feedback review)? Are there obviously missing scenarios?

5. **Iteration potential**: Can this toolkit be upgraded as your skill improves, or is it only useful at your current level?

Please output:
- An overall verdict (can this toolkit genuinely improve practice quality?)
- A rating for each prompt (strong / medium / weak) with specific improvement suggestions
- Issues that must be fixed
- Areas that could be strengthened
- If you were a user of this toolkit, at which stage would you feel 'this prompt isn't helping'?

Yellow placeholders need you to fill in before using the AI.

AI can help you organize ideas, but cannot make final judgments for you. Don't let AI fabricate experiences, cases, or misleading content.

Submit my work