TL;DR. Deliberate practice is not practice with more effort behind it. It is four conditions held at once: one named sub-skill instead of the whole activity, a difficulty just past what you can currently do, feedback fast enough to change the next repetition, and full attention. The worked examples below name the sub-skill and the feedback source in four domains, because "get feedback" is not advice until you say from what. The caveat most articles skip: Ericsson studied fields with clear feedback and centuries of established teaching behind them, and most jobs are not like that.
I have played guitar since I was a teenager. I got noticeably better for about two years, and then I got better at playing the same eight songs for the next fifteen.
That is the normal outcome, not a personal failing. Twenty years of driving does not make anyone a better driver. Repetition makes an activity automatic, and automatic is the opposite of improving, because the point of automatic is that you no longer notice what you are doing.
Deliberate practice is the alternative. It is a specific method with specific conditions, worth being precise about, because the phrase has been diluted into "try harder" almost everywhere it appears.
The four conditions
One sub-skill, not the whole activity. You do not practise guitar. You practise the change from B minor to F, or vibrato at slow tempos, or keeping your picking hand quiet. Naming the sub-skill is most of the work, and it is the step people skip, because "practise guitar for an hour" requires no honesty about what is broken.
A difficulty just past what you can do. If the session feels fluent, you are rehearsing. Correct practice feels slightly awkward and produces visible failures, which is uncomfortable enough that most people drift back to the comfortable version within ten minutes.
Feedback fast enough to change the next repetition. Not feedback next month. Feedback inside the session, so repetition two differs from repetition one. Without it, repetition just laminates the error.
Full attention. Practice with a podcast on is not practice, it is time spent near the instrument. This is the condition I break most often, and the one that quietly wastes the most hours.
"Get feedback" is not advice until you name the source
This is where most writing on the topic stops being useful. Four sources cover nearly everything, and they are the four the deliberate practice planner asks you to pick from before a session starts:
- A coach or skilled peer. The strongest option, because they see what you cannot. Also the rarest.
- A recording of yourself, reviewed critically. The cheapest substitute, and better than people expect. Your ears while playing and your ears on playback are different ears.
- An objective metric. Speed, error rate, split time, hit rate. Weak on nuance, unarguable on direction.
- An expert example to compare against. You produce the thing, hold it next to work you consider good, and find the specific differences.
Pick one before you start. If you cannot name it, the session is not deliberate practice.
Four worked examples
| Domain | What ordinary practice looks like | The sub-skill | The feedback source |
|---|---|---|---|
| Guitar | Playing through songs you know | One chord change, at a tempo just above where it breaks | Phone recording plus a metronome |
| Running a 5k | Running 5k at your usual pace | Holding target pace in the final kilometre, when form goes | Lap splits and cadence from the watch |
| Programming | Shipping features | Forming a correct first hypothesis about a bug | Your own hit rate against the real fix |
| Writing | Writing more | Rewriting the opening paragraph until it earns the second one | An admired original, compared line by line |
Guitar. Set the metronome slightly above the tempo where the change falls apart. Loop only that change for eight minutes, recording it. Listen back once and write down what went wrong: hand arriving late, muted third string, a rushed beat. Next session, that note is the target.
Running. Most amateur running is one steady effort with no feedback until race day. Isolate pace discipline when you are tired: run the first three kilometres easy and treat only the last one as the practice, with the watch as the metric. The split either held or it did not.
Programming. This one is mine, from contract work. Take a bug, write down what you think is wrong before touching any code, then find the real cause. The metric is how often your first hypothesis was right. Mine was worse than I assumed, and the pattern was obvious in writing and invisible in my head: I guessed at the layer I had touched most recently rather than the layer the evidence pointed at.
Writing. Take a paragraph you admire, read it a few times, put it away, then write it from memory and compare the two sentence by sentence. Every difference is either your voice or your weakness, and you usually know which. Benjamin Franklin taught himself this way. The speaking version is the same shape: record a two-minute answer, then count filler words.
The session template
Same shape every time, whatever the domain:
- Warm up on something you can already do. Five to ten minutes. Not practice, just getting the machinery moving so the real work is not spent waking up.
- One target, written as a sentence. "Clean transition into F at 96 bpm." Not "work on chords."
- Short repetitions with a stop after each. The stop is where the correction happens. Continuous playing is where it does not.
- One note on what broke. Three lines at the end: what worked, what broke down, what changes next time.
That last note is the part that compounds, and the part everyone drops first. The planner keeps the reflection next to the session it belongs to, tracks each sub-skill on a 1 to 10 self-rating, and suggests your lowest-rated sub-skill as the next focus, which is a blunt but effective way of stopping you practising your favourite thing forever. There is an optional countdown timer if you want one. Everything stays in your browser.
On length: the planner offers 15 to 90 minutes and flags 45 to 90 as its recommended range. I would treat that as an upper bound rather than a target. My own attention for correction work runs out around 40 minutes, and the tell is that the session stops feeling awkward. For the longer argument about intervals, how long a pomodoro should be covers it, and ultradian rhythms covers the 90-minute claim and how shaky it gets once you are awake.
The 10,000 hours thing, correctly
The 10,000-hour rule is Malcolm Gladwell's, from Outliers in 2008. It is not what the research said.
In the 1993 study of violinists at the Berlin Academy that Gladwell drew on, 10,000 hours was roughly the average accumulated by the strongest group by the age of twenty. It was not a threshold anyone crossed into mastery. Individual totals varied widely, those twenty-year-olds were nowhere near the top of their field, and the hours required differ enormously between domains.
Anders Ericsson, whose work it was, spent much of the rest of his career objecting to the rule. Take that correction if you take nothing else, because the number does real damage: it tells people the input is hours, when the finding was about the kind of hours.
How much practice actually explains
Here I have to be careful, because this is contested and I am not going to pretend otherwise.
The strong reading of the original work, that accumulated deliberate practice accounts for essentially all of the difference between experts and everyone else, has not survived. Later meta-analyses found that practice explains a substantial share of the variation in performance and clearly less than all of it, with the remainder attributed to starting age, working memory, physical attributes and luck of circumstance. How big that share is remains genuinely disputed, partly because studies define deliberate practice differently, and partly because it varies by domain: more of the story in games and music, considerably less in professions where the task itself is not standardised.
What survives every version of the argument is smaller and still useful: how you practise matters more than how much. That is enough to act on. It is not a promise that anyone can become world class, and I would be suspicious of anyone selling you that.
Where the method transfers badly
Ericsson restricted the term to domains with established training methods, objective measures of performance and teachers who already know what good looks like. Music, chess, athletics, some parts of medicine.
Most work is not that. Management is not that. Founding a company is definitely not that. The feedback arrives in quarters, contaminated by luck and market conditions, and nobody can reliably tell you what the correct move was, including afterwards. Running "deliberate practice" on a job like that is mostly theatre.
The honest transfer is narrower. Inside messy work there are sub-skills that do have fast feedback, and those are practisable: writing clearly, debugging, explaining a technical decision out loud, estimating how long something will take. I cannot practise being a founder. I can practise the four or five things that keep costing me when I am bad at them.
The skills you already have are fading meanwhile
Improvement and maintenance are different problems, and the second one stays invisible until you try to use something you have not touched in a year.
The skill decay calculator is a rough instrument for that. You list skills, tag each as motor, cognitive, language or creative, set the last date you practised it, and it estimates how much is left using an exponential curve whose half-life depends on the type and grows more stable each time you log a review. It also generates a spaced review schedule at 1, 3, 7, 14, 30, 60 and 90 days.
Be clear about what that is: a model, not a measurement of you. The decay rates are assumptions baked into the tool. The honest use is as a prompt to go and check whether the language you claim to speak still comes out.
What to do this week
Pick one skill. Name one sub-skill inside it that you are actually bad at. Name the feedback source before you begin. Run three short sessions, write one line after each about what broke, and change one thing between them.
That will tell you more about how you learn than another year of playing your eight songs.
Last updated: December 2026.
Written by Eusebiu, the solo founder building Loggd in public. I build the app in the evenings around contract work and a small daughter, which means my practice budget is about three short sessions a week and I have had to get honest about what fits in them. I share what the aggregate data keeps showing on Threads.
Practice is a habit before it is a method. Loggd tracks your sessions on a forgiving contribution grid, so three short weeks of real work stay visible even when you miss a Thursday. Start free.
