Skill learning: music, sport, coding, drawing
Music, sport, coding and drawing are skills, not syllabuses, and the train-don’t-cover tool applies to all of them, graded or not. The knowledge shape is a list of moves you get wrong, not topics — the triage grid, with moves for rows. When a marker exists, it pays for correctness or fluency on the move actually tested, never for time logged. When no marker exists, you are the marker, checking the attempt against a recording, a model, or whether the thing built actually runs. Either way the mechanism is the same. Practise one part just beyond reach, alone, corrected at once, repeated: a skill grows only at its edge and stalls once the repetition turns comfortable.
Log every attempt against the move it belongs to; the time goes to the move that shows up often and still goes wrong. There is no clock to beat the way a timed paper has one. What matters is how often the edge moves, read across weeks, never inside one sitting. Because a skill rises slowly, run it as a season, checked monthly, not a short crash built around a fixed date. Where the skill chosen is itself a graded test, that entry’s own notice carries the numbers; here the shape is the mechanism, and it does not change with the skill.
Log your next three attempts at the skill, naming the exact move each one got wrong. Pick the move that shows up most and still fails. Drill that one move alone tomorrow, corrected at once, and log the attempt again.
I have put in years on this — I should just be good by now.
Hours alone do not move a skill; only practice at the edge, with feedback, does. Years spent repeating what already feels comfortable teach the hand nothing new, however many hours they add up to. Log the move that goes wrong, drill that alone, and check the loop closes before counting the hour at all.