Within one pattern category, why sequence easy-then-medium instead of hopping categories daily?

## The claim, stated precisely Given a fixed number of problems, the *order* you attempt them in changes what you retain. Grouping consecutive attempts within one technique category, ascending in difficulty, produces transferable patterns. Shuffling categories day to day produces recognition of specific problems. Same problems, same hours, different outcome. ## Why grouping produces abstraction A pattern is an abstraction, and abstraction is built by comparison. When you solve one problem you have a single data point: a sequence of decisions that happened to work. You cannot tell which decisions were essential to the technique and which were incidental to that problem's dressing. Solve a second problem in the same category immediately after and the comparison becomes unavoidable. The window still expands and contracts, but now the shrink condition is different. The tree recursion still returns a value upward, but now the value being combined is a pair, not a number. The moment you notice *the same skeleton with a different filling*, you have extracted the skeleton — and the skeleton is the thing that survives contact with a problem you have never seen. Spread those two problems three weeks apart and the comparison never happens, because by the time you meet the second you have forgotten the shape of the first well enough that it feels like a fresh problem. You solve it again from scratch, and store a second anecdote. ## Why easy comes before medium inside the category An easy problem in a category is usually the technique with the twist stripped out. It teaches the mechanism in isolation: what the invariant is, what state you carry, why the loop terminates, which boundaries are dangerous. That gives you a correct baseline. A medium in the same category is generally the same mechanism plus one complication — a second constraint, a different comparison key, an extra dimension of state. When you already hold the baseline, the complication is the only unknown, and your attention goes to exactly the part that is new. That is efficient learning. Start the category on the medium and you are debugging two unknowns at once: is my understanding of the technique wrong, or am I mishandling the twist? You cannot tell, and the usual outcome is reading a solution, which teaches recognition rather than derivation. This is also why curated lists that order problems within a category are more useful than lists that only group them. The grouping gives you the comparison; the ordering gives you the clean baseline before the complication. ## What daily hopping actually optimizes Hopping categories every day is not irrational — it optimizes for two real things: motivation (variety is more pleasant than repetition) and simulated interview conditions (a real round gives you a problem with no category label, so practising cold identification has value). But it buys those at the cost of abstraction. If the first problem you meet in a category is also the only one you meet that week, every attempt is a cold start, and cold starts are where you are most likely to consult a solution. Consulting a solution produces the strongest form of the trivia failure: you now recognize *that* problem instantly and are no better at its siblings. The good compromise is two-phase. Learn in blocks — a category at a time, easy first, until the skeleton is stateable in a sentence. Then, separately, take mixed unlabeled problems to practise identification, which is the skill hopping was reaching for. Mixing is a *testing* mode; blocking is a *learning* mode, and using the testing mode to learn is what produces trivia. ## The costs of blocking, and how to bound them Blocking has genuine downsides and a plan should acknowledge them: - **Staleness.** A category finished in week one and untouched for a month decays. Blocks need revisits, not just completion. - **False fluency.** Inside a block you know the category before you read the problem, so identification is free — which is exactly the skill the interview tests and the block does not train. - **Diminishing returns.** After three or four problems the category's shared structure is usually visible; further problems in the same block mostly teach that category's rarer sub-cases and are better spent opening the next blank category. - **Boredom is a real constraint.** A plan you abandon in week three is worse than a slightly suboptimal plan you finish. A workable shape: blocks of three to four problems per category, easy then medium, every category opened before any category is deepened, then a mixed unlabeled set that revisits older categories and restores what has gone stale. ## The failure to recognize in yourself The diagnostic is simple. Pick a category you practised. Without opening anything, state in one sentence what the technique maintains and when it applies. If you can, the block worked. If instead you can only recall specific problems you solved, you collected trivia — and more problems, in that same shuffled order, will collect more of it.