Tutorials feel like progress. This is a workshop for the part that comes after: seven 5–15 minute apprenticeships where you make something someone could see, check it, and try again — with an honest account of what practice can and can't do.
What this is: free exercises, worked examples and a reading list you can use on your own, today, with a pen and a browser.
What this is not: a course, a certificate, a skills score or a claim that tools, teachers and schools don't matter. Nothing you type here leaves your device.
Seven notebooks, one bench. Pick any notebook to begin — the little pencil walks to wherever you're up to.
Step one
Choose a path
Every apprenticeship stands on its own. If you'd like a suggested order, pick what you're after. You can change your mind at any time.
Your path
About saving: your chosen path and which notebooks you've marked as tried are kept only in this browser (localStorage, key apprentice-path-v1). Nothing is sent anywhere. Anything you type into exercises is never saved.
The workshop
The seven apprenticeships
Each notebook has the same four pages: a learning goal, an explicit example with the reasoning shown, a 5–15 minute exercise and an honest reflection — including what the exercise can't teach you. A pen and paper beat any form here.
1
Choose a good question
≈ 10 minutes · pen and paper
On your path
Learning goal
Turn a big, vague interest into a working question you could make progress on this week — without losing the reason you cared.
Vague questions ("How do I get better at writing?") can't be answered by anything, so they tend to be answered by more content. A working question can be answered by doing something.
Example Illustration
Start “How do I get better at design?” — real, but nothing you could do at 4 pm today.
Narrow the object “How do I make my one-page portfolio easier to read?” — now there's a thing to look at.
Make it observable “Can a friend find my email address on it in under ten seconds?” — now you'd recognise an answer.
Point at an action “What on the page competes with the email address for attention, and what happens if I remove two of those things?” — now the answer changes what you do.
Each rung changes one thing. Changing everything at once usually produces a different question, not a better one.
Exercise
2 min Write a big question you genuinely care about. Don't make it impressive.
5 min Rewrite it three times, changing only one thing each time: a narrower object, an observable result, a next action.
2 min Run the five tests below on each version. Cross out any version that fails two.
1 min Circle one. Write the first fifteen-minute step underneath it.
Could I start on it today?
Would I recognise an answer if I saw one?
Would the answer change what I do?
Is it mine — not borrowed from a headline?
Is it small enough to be wrong cheaply?
Honest reflection
Did your question get smaller and better, or smaller and duller? Narrowing sometimes kills the thing you cared about. If so, keep the big question as a compass and the small one as a step; you need both.
What this can't teach you: whether your question matters to anyone else, or whether it has already been answered well. For that you need reading and other people.
2
Write a strong paragraph
≈ 12 minutes · a topic you know from experience
On your path
Learning goal
Write one paragraph that makes one claim a reasonable person could disagree with, grounds it in something concrete, and says what follows.
Example Illustration
Before
Learning is very important in today's fast-changing world. There are many ways to learn, such as videos, books and courses. It is important to choose the right one and stay consistent. Consistency is the key to success.
After
Most tutorials are designed to be finished, not used. A twelve-minute video on spreadsheet formulas ends when the presenter's example works; it rarely asks you to break the formula, guess why, and fix it with your own numbers. That missing step is where the skill lives. So before the next tutorial, decide what you'll make with it in the ten minutes after it ends — and if you can't name anything, the video is probably entertainment, which is allowed, as long as you call it that.
Why the second works: the “before” has four sentences and no claim anyone would argue with — every line would fit any essay on any topic. The “after” makes a disputable claim (sentence 1), shows a specific case (2), names the consequence (3) and earns a practical ending that even admits an exception (4).
Exercise
2 min Pick something you know from lived experience — a job, a commute, a family recipe. Write your claim as one sentence someone could disagree with.
4 min Write the most concrete support you have: a moment, a place, a number you actually know.
2 min Write what follows if you're right.
4 min Read it aloud. Delete every sentence that could appear in an essay about anything.
Honest reflection
Which sentence did you most want to keep that you cut? That sentence usually felt safe. Notice what safety cost.
What this can't teach you: a clear paragraph can still be wrong. Clarity just makes the error easier to find — which is the point. Long-form structure, voice and research are separate apprenticeships.
3
Critique an AI answer
≈ 12 minutes · any AI tool, or the sample below
On your path
Learning goal
Read a fluent machine-written answer the way a careful editor would: separate what's checkable, what's vague, what's wrong and what's missing.
Example Fiction
This answer was written for this page to imitate common patterns. It is not output from any named product. Mark each sentence yourself before opening the annotated version.
Prompt: “What's the best way to learn to code as a beginner?”
“The best way to learn coding is to practise for 10,000 hours, as research has proven this is what it takes to master any skill.”
“Start with Python because it is the easiest language and is used by 90% of companies.”
“Watch coding tutorials every day and you'll be job-ready in three months.”
“Remember, consistency is key!”
Lift the flap: an annotated critique
“10,000 hours… proven… any skill.” Overstated. The 1993 deliberate-practice study looked at musicians and did not establish a universal number; a 2014 meta-analysis found practice explained very different shares of performance in different fields, and very little in professions (see chapter 5). “Proven” is doing illegitimate work.
“Easiest… 90% of companies.” “Easiest” depends on what you want to build. The statistic arrives with no source; a precise-sounding number without a source is a reason to ask, not to believe. Python is still a reasonable, common first choice — the advice isn't wrong, the justification is.
“Watch tutorials… job-ready in three months.” Treats watching as practice, and promises an outcome with no conditions: hours per day, prior experience, which job, which local market.
“Consistency is key!” Not wrong, not useful. It can't be checked and doesn't tell you what to do on Tuesday.
What's missing: the answer never asked what you want to make. It never mentions feedback, debugging, or reading other people's code. A good human mentor would have asked a question back.
Exercise
2 min Ask any AI tool a question in an area you know well — your job, your neighbourhood, your hobby.
6 min Mark each sentence: ✓ checked, ? checkable but unchecked, ~ vague, ✗ wrong as far as you know. Check at least one “?” against a real source.
2 min Write the one question the answer should have asked you before answering.
2 min Ask that question back in the chat and compare the new answer.
Honest reflection
Was it easier to find problems because you chose your own area? That's the uncomfortable lesson: you can audit fluent answers mostly where you already know something. Basic knowledge doesn't become pointless when tools get good; it becomes the thing that lets you use them.
What this can't teach you: professional fact-checking. That needs time, sources and sometimes people on the phone. Four marks on a page are a habit, not a method.
4
Fix a broken simple webpage
≈ 15 minutes · works best on a laptop; possible on a phone
On your path
Learning goal
Find and fix eight common problems in a tiny web page, and say for each one who it hurts — a phone user, a screen-reader user, a search engine or someone in bright sunlight.
Example: one bug, explained Fiction
The page below belongs to an invented tiffin service. Its photo tag is <img src="tiffin.jpg"> with no alt text. If the image fails to load — and here it will, because the file doesn't exist — sighted visitors see a blank, and a screen-reader user hears little or nothing useful. The fix is a short description of what the picture tells someone: alt="Steel tiffin with rajma, rice and two rotis". Notice the reasoning: not “add an alt attribute because a checker said so”, but “what would someone miss without it?”
Exercise
Edit the code, then preview it and run the checks. There are eight problems; the checker tests for each. Your edits are not saved.
Preview (scripts disabled, sandboxed)
Checks
Press “Check my page” to run eight simple checks.
Hints: the eight problems
No <!doctype html> — browsers fall back to “quirks mode”.
No lang on <html> — screen readers may guess the pronunciation wrong.
No viewport meta — phones render it zoomed-out and tiny.
The <h1> is never closed, so the paragraph becomes part of the heading.
The image has no alt text.
The headings jump from h1 to h3; outlines and screen-reader navigation get confusing.
“Order now” links nowhere (href="") — the most important link is dead.
Light grey #ccc on white is very hard to read, and the deadline is important information.
Honest reflection
Which fixes changed what you could see, and which only changed what a phone, a screen reader or a search engine would experience? Most beginners fix the visible ones first. That's a fact about attention, not about importance.
What this can't teach you: passing eight checks doesn't make a page accessible or good. The checker here is deliberately simple (the contrast check just looks for that grey). Real accessibility work involves testing with people and proper tools — and for a business site, it's reasonable to want a practitioner's eye.
5
Run a fair micro-observation
≈ 15 minutes to set up · then a few days of noting
On your path
Learning goal
Test a small hunch about your own life without fooling yourself — by deciding what counts before you look.
Example Illustration
Hunch: “I get more done on days I don't check messages before 10 am.”
Unfair version
Try it on a day you're already excited. Afterwards, remember how good it felt. Quietly forget the Thursday it didn't work because “that day was unusual”.
Fairer version
Decide first: “done” means tasks finished from a list written the night before. Alternate days — A, B, A, B — over six workdays, decided in advance. Record every day, including the bad ones, and note anything else that changed (sleep, deadlines, illness).
A second, hypothetical case from shop life: someone who believes rainy days bring fewer customers could write the weather and a simple count in a notebook every evening for a few weeks — before deciding — instead of relying on the rainy days that felt quiet. Hypothetical
Six days can't prove anything. They can make you less wrong about yourself — which is more than memory usually manages.
Exercise
3 min Write your hunch in one sentence.
3 min Write what result would make you drop it, and what would make you keep it. If nothing could change your mind, it's a belief, not a hunch — fine, but don't test it.
4 min Define the count so a stranger could do it: what exactly gets a tally mark?
3 min Fix the schedule in advance: which days or hours are A, which are B.
2 min Decide where you'll record it. Observe yourself or anonymous counts — never note who strangers are.
Pocket tally (this visit only — closes when you leave the page)
With tiny numbers, a difference of a few can easily be chance or circumstance. This tally counts; it doesn't judge.
Honest reflection
Did you want a particular result? Where could that wanting have leaked in — the definition, the schedule, the excuses? Writing your “drop it if…” sentence first is the single best defence.
What this can't teach you: statistics, causation or anything about other people. A notebook experiment on yourself is a way to notice; research needs design, samples and ethics review.
6
Describe a scent
≈ 10 minutes · three things from a kitchen
On your path
Learning goal
Build a smell vocabulary through comparison and layers, instead of reaching for “nice”, “strong” or “fresh”.
Safety first: use whole kitchen materials — spices, citrus peel, tea leaves, coffee. Don't sniff essential oils or aroma chemicals straight from the bottle, don't put anything on your skin for this exercise, take fresh-air breaks, and skip it if smells give you headaches or you have breathing sensitivities.
Example: a crushed green cardamom pod Illustration
1 · What it is
Green cardamom, pod cracked between two fingers.
2 · What it's like
Cool and eucalyptus-like at first, a little camphor; then lemony; closer to the husk, something sweeter and softer.
3 · How it moves
Loud for the first breath, then the cool part fades in about a minute and the sweet, woody side stays.
4 · What it means to me
“Afternoon chai.” Kept separate, because it belongs to the smeller, not the smell. Someone else's layer 4 will differ; their layers 2 and 3 should overlap with yours.
Evidence Describing smells is famously hard for English speakers — but perhaps not for everyone. In a small study, speakers of Jahai (a language of the Malay Peninsula with a rich odour vocabulary) named odours about as consistently as colours, while the English-speaking comparison group found odours much harder (Majid & Burenhult, 2014). Interpretation Part of the difficulty may be language and practice, not only the nose. One small study; it doesn't show that training makes anyone a perfumer.
Exercise
1 min Choose three materials that are similar — say, cinnamon, clove and nutmeg; or three teas. Similar is harder and teaches more.
2 min each Write all four layers for each. In layer 2, compare to at least two other things.
3 min Ask someone to hand them to you, eyes closed, in a random order. Can your own notes tell them apart?
Honest reflection
Which layer was hardest? For most people it's layer 3, because noticing change means smelling more than once. Did your words survive the blind test, or did you recognise them by something your words never captured?
What this can't teach you: describing smells is not the same as composing with them. Perfumery involves material knowledge, dilution, safety and regulation — areas where a teacher, not a webpage, is the right guide.
7
Give constructive feedback
≈ 12 minutes · one willing person and one piece of their work
On your path
Learning goal
Give feedback someone can act on — about the work, not a verdict on the person — and leave the decision with them.
Evidence In a large review of feedback studies, feedback improved performance on average, but over a third of the feedback interventions decreased performance (Kluger & DeNisi, 1996). Interpretation Their theory proposed that feedback is less helpful when it pulls attention away from the task and towards the self. So: describe the work, not the worker.
Example Illustration
A friend shares the “after” paragraph from apprenticeship 2.
Less useful
“This is great!” “Bit boring tbh.” “You're a natural writer.”
More useful
Observation: “Your second sentence, about the twelve-minute video, is where I started to believe you.” Effect: “Before that I was nodding, not convinced.” Question: “Could the first sentence be as specific as the second? What would you lose?”
All three “less useful” lines are about the person or a feeling with no handle. The better note quotes the work, reports its effect on one reader, and asks rather than instructs.
Exercise
1 min Ask someone for a piece of work they're happy to share. Don't send unrequested critique to strangers.
1 min Ask: “What kind of feedback would help right now?” Then honour the answer.
7 min Write three notes, each as observation → effect → question or suggestion.
3 min Cut the least useful. Deliver at most two. End by giving the decision back: “Your call.”
Honest reflection
Did they use it? If not, was your note unclear, unwanted or simply wrong? All three happen to everyone. And notice how it feels to receive notes this way — next time, ask for feedback in this format.
What this can't teach you: a peer's feedback reports a reader's experience; an expert's can also show the craft's options and standards. You need both, and they aren't interchangeable.
The argument
Output is not expertise — but making things is still the best teacher most of us can reach
Nine short chapters on tutorials, practice, feedback, revision and transfer: what the research supports, where it's oversold, and where you should stop teaching yourself and find a person.
1 The tutorial trap
Interpretation
Somewhere right now, someone is on their fourth tutorial about the same thing, and each one has felt like progress. The feeling is real. Watching a skilled person work is genuinely useful: you see the order of moves, the vocabulary, what “good” looks like. For a beginner, a clear worked example can be worth more than an hour of fumbling.
The trap is that familiarity feels like competence. You recognise every step as it happens, so you assume you could produce the steps. Then you open a blank file, or pick up the knife, or face the client, and the steps aren't there. Recognition and production are different abilities, and tutorials mostly train the first.
Most tutorials are also designed to end — at the moment the presenter's example works. That is precisely the moment when learning would start: when you try it on your own material, it breaks, and you have to work out why. The answer to the tutorial trap is not to stop watching. It's to attach every piece of consumption to three things: something observable you'll make, a way to find out how it went, and a second attempt.
2 Five verbs: consume, practise, get feedback, revise, transfer
InterpretationIllustration
What each verb leaves behind
Verb
What you can observe afterwards
Typical trap
Consume
Notes, a sense of the shape of the thing
Mistaking recognition for ability
Practise
A made thing, plus a list of where you got stuck
Repeating what you can already do
Get feedback
Specific information about the gap
Asking for praise; receiving verdicts
Revise
Version two, visibly different from one
Polishing the surface, not the problem
Transfer
The skill working on a new problem
Assuming it will transfer by itself
A worked example. Two people want to lay out a simple web page. Person A watches three videos on CSS layout. Each is clear; by the third, everything feels obvious. They open an empty file and can't remember which property goes on the parent. Person B watches one video, closes it, and rebuilds the example from memory — which takes four tries and one peek. They show it to a friend on a phone, who can't find the button. They change the spacing and the order. A week later they lay out a menu card for a relative's stall, a problem the video never mentioned, and it half-works.
Person B spent perhaps twice as long in the first week. What they gained wasn't mainly more knowledge; it was evidence of what they couldn't yet do, which is the only thing you can practise against. That last step — the menu card — is transfer, and it's the least automatic of the five. Researchers have argued for decades about how often learning carries over to new contexts; Barnett and Ceci (2002) proposed a taxonomy partly because studies defined “near” and “far” transfer so differently. The practical lesson is modest: if you want a skill to travel, practise it in more than one setting on purpose.
3 Scaffolding — and taking it down
Established frameworkInterpretation
The word comes from a 1976 paper by David Wood, Jerome Bruner and Gail Ross, who studied a tutor helping three-, four- and five-year-olds build a pyramid from interlocking wooden blocks. The tutor didn't do the task. She handled the parts the child couldn't yet manage: keeping the goal in view, simplifying the task, pointing out what mattered, managing frustration and demonstrating when needed — and left the rest to the child.
The metaphor contains its own instruction: a scaffold is meant to come down. Help should shrink as the learner succeeds and return when they struggle. That's the part most often lost.
Modern tools are extraordinary scaffolds. Templates, autocomplete, code assistants and chatbots can hold a whole task together for a beginner. The risk is not that they help; it's that a scaffold you never remove becomes the building. Two quick tests: Could I make a smaller version of this with the tool switched off? and Could I explain what the tool did and why? If the answer to both is no, you've produced output. That may be exactly what you needed — but call it output.
Limit: some scaffolds should stay up for life. Nobody needs to give up spell-check, calculators or a pilot's checklist to prove a point. The real question is narrower: which parts of this skill do you want to own, because you'll need to judge, adapt or teach them?
4 Retrieval: the test is part of the lesson
EvidenceInterpretation
In experiments reported by Henry Roediger and Jeffrey Karpicke (2006), students read short prose passages and then either studied them again or took recall tests on them. On a test five minutes later, re-studying did better. On tests two days and a week later, the students who had practised recalling remembered more.
The interesting part is the reversal. Re-reading produces a pleasant sense of fluency that pays off in the very short term and misleads you about the long term. Pulling something out of your head is harder, feels worse, and works better for keeping it. In practice: close the tab and reproduce the idea, the code, the recipe — then check.
Limits: retrieval strengthens whatever you retrieve, including mistakes, so it needs a check against the source. Much of the research concerns remembering text and facts; judgement, taste and physical skill need more than recall. And the effort is not a sign it's failing — though if it's all effort and no success, the task is probably too far ahead of you and needs a scaffold.
5 Deliberate practice: a useful idea with an oversold number
Established frameworkEvidence
Anders Ericsson, Ralf Krampe and Clemens Tesch-Römer (1993) studied violinists and pianists and argued that expert performance is closely tied to accumulated deliberate practice: effortful activity designed specifically to improve, usually with clear goals and feedback, often structured by a teacher — as distinct from playing for fun or performing. The popular “10,000 hours” rule flattened this into a number and a promise the paper did not make.
In 2014, Brooke Macnamara, David Hambrick and Frederick Oswald pooled studies in a meta-analysis. Deliberate practice explained about 26% of the variance in performance in games, 21% in music, 18% in sports, 4% in education and less than 1% in professions. These estimates were contested — partly over how “deliberate practice” was defined in the included studies — but the broad picture is hard to dismiss: practice matters, it matters more in some fields than others, and it is not the whole story.
Games≈26%
Music≈21%
Sports≈18%
Education≈4%
Professions<1%
Share of performance differences explained by deliberate practice, by domain, as reported by Macnamara, Hambrick & Oswald (2014). Bars are drawn to scale out of 100%. Reported summary figures, not this site's data.
Interpretation Look at where practice explains most: domains with clear rules and fast, unambiguous feedback. In messy work, it's hard to know what to practise and the feedback arrives late and noisy. The lesson for an ordinary learner isn't “practice doesn't work”. It's: where feedback is slow, build small, fast feedback loops yourself. That's all the seven apprenticeships are. And remember the other things that explain performance — teaching quality, time, health, money, starting point, temperament, luck — before you blame yourself for a plateau.
6 Turning passive content into observable skill
InterpretationIllustration
An “observable skill” is one another person could see, hear, read, run, taste or smell. Here is a recipe for converting anything you've watched or read:
Name an output. What would exist afterwards that didn't before?
Shrink it to a version you can make in fifteen minutes.
Make it with the source closed. Peek only when stuck, and note where.
Compare against a standard: the source, a good example, a checklist or a person.
Make version two. Change one thing on purpose.
Move it to a situation the source never mentioned.
Worked example. You listen to a podcast about negotiating. Output: two sentences you would actually say to ask for a deadline extension. Fifteen-minute version: write them, say them aloud twice without notes. Compare: ask a friend to play your manager and tell you which sentence made them want to say no. Version two: replace the apology with the reason. Move it: use the same structure, weeks later, to ask a landlord to fix a tap.
Try it: the content-to-practice converter
Nothing you type is saved or sent; it's only used to fill in the steps below.
7 When you need an expert, not another tutorial
Established frameworkInterpretation
Self-directed doesn't mean unguided. Paul Kirschner, John Sweller and Richard Clark (2006) argued that minimally guided instruction works poorly for novices compared with explicit guidance such as worked examples. Cindy Hmelo-Silver, Ravit Golan Duncan and Clark Chinn (2007) replied that well-designed problem-based and inquiry learning is heavily scaffolded, not minimally guided. The debate is real; both sides agree that beginners left alone with a hard problem tend to struggle.
Look for a teacher, mentor or professional when:
Mistakes are dangerous or expensive — wiring, medicine, money, food safety, concentrated aroma materials, legal documents.
You can't yet perceive your own errors — pronunciation, posture, security holes in code, tone in a hard email.
The knowledge is tacit — it lives in someone's hands and judgement and doesn't fit in a video.
Practice might be ingraining a fault — the more you repeat a bad grip, the harder it is to change.
You've plateaued — and your own feedback loops have stopped telling you anything new.
AI tools sit in an odd place here. They can explain patiently at 2 am, and they can be confidently wrong, and they can't watch you hold the knife. Treat them as a well-read study partner, not as the expert in the room — which is exactly why apprenticeship 3 exists.
8 The strong objection: “learn by doing” is a story told by people who could afford to
Interpretation
“Free tutorials and AI tools don't make learning equal. They reward people who already have time, a device, reliable data, fluent English, a quiet room and someone to call when they're stuck. Telling everyone to ‘just make things’ dresses up privilege as method.”
Most of this is right. Access to time — free of paid work and care work — is unevenly distributed. So are mentors, permission to fail in public, and the confidence to show unfinished work. Cheap tools lower some barriers and raise others: when anyone can produce something that looks finished, the bar for being taken seriously rises. And output is not expertise. A polished website, essay or logo is output. Expertise is the ability to make good decisions when the situation changes, explain why, and notice what's wrong before anyone else does.
This page makes a narrower claim than the objection attacks: whatever time and access you have, small observable experiments with feedback get more learning out of it than consumption alone. It does not claim effort overcomes structure. It also implies a demand on institutions — schools, employers, platforms — to provide what self-teaching can't: feedback, mentorship, safe chances to fail and time. That argument is developed further on Human Capital for Humans.
Counterexamples worth keeping. Some people learn deeply from reading alone, especially where understanding is the goal — history, philosophy, mathematics read slowly with a pencil. (Arguably that reading is making: an argument rebuilt in your head.) And some people make things prolifically and barely improve, because they never seek feedback or revise. Quantity without revision is just output with a timestamp.
9 Joy, rest and the right to be a beginner
Interpretation
Everything above can be turned into a stick to beat yourself with: more reps, more feedback, more optimising. Please don't. Tired practice tends to be worse practice. Not every interest needs to become a skill tree, and a hobby you are merely bad and happy at is a complete thing.
The apprenticeships are 5–15 minutes on purpose. The map ends at a bench. If you notice a gap between the learner you'd like to be and what you actually do on a Tuesday evening, that gap is interesting — not a moral failing. Sahil explores it as a proposed inquiry, not a law, in The Identity–Action Gap. Often the fix is smaller friction, not more willpower: a pen already on the notebook, a fifteen-minute version, a friend who's expecting to see version two.
Common mix-ups
Six misconceptions
“I understood the video, so I've learned it.”
You've learned to recognise it. Production is a separate skill; test it by making something with the video closed.
“If it feels hard, I'm bad at this.”
Retrieval and practice often feel worse than re-reading while working better. Difficulty that ends in some success is usually the good kind.
“10,000 hours makes anyone an expert.”
No study established a universal number, and practice explains very different shares of performance in different fields.
“Self-taught means taught by no one.”
Self-directed learners use books, examples, communities and mentors constantly. The self part is the steering, not the isolation.
“More output means more skill.”
Only if output is followed by feedback and revision. Ten unexamined attempts can teach less than two examined ones.
“AI makes learning basics pointless.”
Basics are what let you judge AI output. You can only audit a fluent answer where you already know something.
The open question
If a tool can produce the output, which part of the skill is worth keeping in your own hands — and who gets to decide that: you, your employer, or the tool's maker?
There isn't a general answer, and it will be different for a nurse, a novelist and a shopkeeper. If you have a thoughtful one, Sahil would like to hear it. Longer essays on learning, work and identity are collected on his Medium archive.
Take-home worksheet
One page to print or copy into a notebook. The print button prints only this sheet, with your chosen path listed.
If this page is about what one learner can do, that essay asks what organisations owe learners: capability, dignity and time — the support self-teaching can't supply on its own.
Why “I'm someone who practises” and actually practising on Tuesday can drift apart — explored there as a proposed inquiry, with a friction map that's labelled as illustrative.
The longer-form writing archive, for when you'd like to read slowly with a pen — which, per chapter 8, also counts as making.
Check the workings
Sources, and what they do — and don't — show
Bibliographic details were checked against Crossref records or the publisher's copy in October 2026. Summaries are this page's reading; follow the links to judge for yourself.
Wood, D., Bruner, J. S., & Ross, G. (1976). The role of tutoring in problem solving. Journal of Child Psychology and Psychiatry, 17(2), 89–100. doi:10.1111/j.1469-7610.1976.tb00381.x
Establishes: the origin of “scaffolding” and the functions a tutor served for young children on a block-building task. Doesn't: show that any particular modern tool or course is good scaffolding, or describe adult self-study.
Roediger, H. L., III, & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249–255. doi:10.1111/j.1467-9280.2006.01693.x
Establishes: in these experiments, practising recall of prose passages beat re-studying on delayed tests, though re-studying won at five minutes. Doesn't: show that retrieval alone builds judgement, taste or physical skill.
Ericsson, K. A., Krampe, R. T., & Tesch-Römer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100(3), 363–406. doi:10.1037/0033-295X.100.3.363
Establishes: the deliberate-practice framework and its association with performance level among musicians. Doesn't: establish a universal “10,000 hours” rule or that practice is the only cause of expertise.
Macnamara, B. N., Hambrick, D. Z., & Oswald, F. L. (2014). Deliberate practice and performance in music, games, sports, education, and professions: A meta-analysis. Psychological Science, 25(8), 1608–1618. doi:10.1177/0956797614535810
Establishes: pooled estimates of how much performance variance deliberate practice explained, varying strongly by domain. Doesn't: settle the definitional dispute about “deliberate practice”, or say practice is useless where estimates are low.
Kluger, A. N., & DeNisi, A. (1996). The effects of feedback interventions on performance: A historical review, a meta-analysis, and a preliminary feedback intervention theory. Psychological Bulletin, 119(2), 254–284. doi:10.1037/0033-2909.119.2.254
Establishes: feedback helped on average, yet over a third of interventions reduced performance; proposes that feedback drawing attention to the self is less helpful. Doesn't: validate the specific observation–effect–question format used here (that's a practical convention).
Kirschner, P. A., Sweller, J., & Clark, R. E. (2006). Why minimal guidance during instruction does not work. Educational Psychologist, 41(2), 75–86. doi:10.1207/s15326985ep4102_1 — with the reply: Hmelo-Silver, C. E., Duncan, R. G., & Chinn, C. A. (2007). Scaffolding and achievement in problem-based and inquiry learning. Educational Psychologist, 42(2), 99–107. doi:10.1080/00461520701263368
Establishes: a live, well-argued debate on how much guidance novices need. Doesn't: resolve it — read both before quoting either.
Barnett, S. M., & Ceci, S. J. (2002). When and where do we apply what we learn? A taxonomy for far transfer. Psychological Bulletin, 128(4), 612–637. doi:10.1037/0033-2909.128.4.612
Establishes: a framework for describing how far learning has to travel (content and context dimensions). Doesn't: give a recipe guaranteeing transfer.
Majid, A., & Burenhult, N. (2014). Odors are expressible in language, as long as you speak the right language. Cognition, 130(2), 266–270. doi:10.1016/j.cognition.2013.11.004
Establishes: in a small comparison, Jahai speakers named odours about as consistently as colours, unlike English speakers. Doesn't: show that vocabulary exercises improve smelling ability or perfumery skill.
This page was written and built with the help of an AI assistant. All visuals are original SVG and CSS. The tiffin service, the AI answer and all named examples are fictional or illustrative.
Thanks for walking the path
Pick one notebook, make one small thing, and show it to one person. Then rest.