Practice routes, taught properly.

Discover looks like a lookup table, and that is exactly why most people learn nothing from it. You pick a key, you read the roman numerals, you nod, you leave. Reading is not practice. The mode becomes a practice route the moment you make it a call-and-check loop: you predict, then you press, then you compare. Prediction is what forces the knowledge out of the screen and into your hands.

The unit of work here is the key, not the chord. A key is seven chords that share a neighbourhood: three major, three minor, one diminished, in a fixed order — I, ii, iii, IV, V, vi, vii°. That order never changes in any major key. Once you can recite the pattern of qualities without the key attached, you own every major key at once, and Discover stops being a reference and becomes a test you can grade yourself on in ten seconds.

Then add sound to sight. Every time you reveal a chord, play it, hold it, and say out loud what it does: home, away, tension, resolution, sad-relative, unstable. Function is easier to remember than spelling, and it is the thing you actually reach for when you are writing. Nobody sits at a keyboard thinking 'I need the third diatonic triad'. They think 'I need the one that sounds like a doorway'.

Finish by breaking the key. Play the diatonic seven, then borrow one chord from the parallel minor — usually iv or ♭VII — and hear how much colour a single outside chord buys you. That contrast is the reason to learn the inside of the key in the first place.

Genres are not defined by chords alone, but chord habits are one of the loudest signals a listener hears in the first two bars. Genre Matcher exists to make that signal conscious. Build a loop, see which styles use it, and you find out what your instincts already lean toward — usually something more predictable than you would like to admit.

Run it forwards first: build the progression you naturally reach for, and read the genres it returns. If the answer is always the same two, you have found your default. A default is fine for finishing songs and fatal for writing a varied record.

Then run it backwards, which is where the real learning is. Choose a genre you never write in — gospel, bossa, nineties country, shoegaze — and let the mode hand you its habits. Play them badly for a while. Borrowing a progression from a style you do not live in is the fastest way to break out of your loop, and because you are only borrowing harmony, the result still sounds like you.

The last step is deliberate mismatch. Take a genre progression and play it at the wrong tempo, in the wrong register, with the wrong rhythm. Most memorable songs live in that mismatch: a country turnaround at half speed with sustained pads stops being country and starts being yours.

The Daily Puzzle is the smallest practice unit in the app, which is what makes it the only one people actually keep. A single puzzle a day beats a two-hour session every third week, because recall improves with the number of separate occasions you retrieve something, not with the total minutes spent staring at it.

The streak is a hook, not a goal. What makes the day useful is the thirty seconds after the puzzle: write down what the puzzle asked, what you answered, and — if you were wrong — the specific reason. Not 'I got it wrong'. Something like 'I heard the third as major because the melody sat on the fifth'. Reasons repeat. When the same reason shows up three days running you have found your actual weak spot, and that is what Focus Mode is for.

Speed is worth nothing here. If you can solve the puzzle instantly, slow down and justify the answer out loud before submitting; if you cannot justify it, you guessed correctly, which teaches nothing. Guessing right is the most expensive outcome in ear training because it feels like learning.

Protect the streak with a floor, not a ceiling. On a bad day the deal is one puzzle, one line in the log, done. Never let the ritual become something you can fail.

Test Knowledge is adaptive, which means it hunts. Answer well and it climbs until you fail; that failure point is the whole product. The number at the end matters far less than the shape of the run — where it pushed, where you stalled, and what kind of question broke the streak.

Take the quiz cold. A warmed-up ear and a freshly-read cheat sheet produce a flattering score and a useless diagnosis. Cold results tell you what is durable, and durable knowledge is what you have while writing at eleven at night.

Sit with the wrong answers rather than retaking immediately. For each miss, do three things: state the correct answer, state why your answer was plausible, and play both on the keyboard so your hands hear the difference. The second step is the one everyone skips and the one that stops the same error recurring — wrong answers are rarely random, they are near-misses with a logic of their own.

Retake weekly, not daily. Adaptive tests reward familiarity with the item pool, so frequent retakes measure your memory of the questions rather than your understanding. A weekly cadence keeps the signal clean, and a fortnightly climb of even one difficulty band is real progress.

Focus Mode is deliberately unpleasant. It weights the puzzle stream toward whatever you are worst at, so your hit rate drops and stays down. That is not a bug: practice that feels easy is practice happening inside what you already know. The zone worth living in is roughly seventy to eighty per cent correct — high enough to keep going, low enough that something is being repaired.

Run it in short, dense blocks. Ten to fifteen minutes of weighted puzzles will do more than an hour of mixed practice, and past twenty minutes accuracy falls for reasons of attention rather than knowledge, which pollutes the weighting. Stop while you still care about the answers.

Use the playable reveal every single time you miss. When your attempts run out, the correct chord or scale lights up on the keyboard and stays playable — do not read it and move on. Play it three times, sing the top note, then play the thing you guessed and hear the gap. A miss you correct with your hands is worth ten misses you correct with your eyes.

Then narrow. If the stream shows you keep failing minor sevenths against a major backdrop, stop taking whatever comes and hand-weight the mode to that one item for a week. Broad practice maintains; narrow practice repairs. Alternate them — one narrow week, one broad week — and neither the weak spot nor the general fluency is left behind.

Every other mode asks 'what is this?'. Chord Detective asks 'what could this be, and what would rule the others out?'. That is a different skill and it is the one you use in the real world, where you almost never hear a chord in isolation — you hear a bass note, a melody, and a vague colour, and you have to reason your way to the name.

Work the clues like a process of elimination rather than a guessing game. A clue narrows the field; it rarely names the answer. Before you touch the keyboard, say out loud what the current clue removes. 'Contains a minor third' removes every major and suspended chord. If you cannot say what a clue eliminates, you have not used it — you have just read it.

Keep a shortlist of two or three candidates and test them, rather than committing to a favourite and looking for support. The classic failure here is confirmation: you decide it is a minor seventh in bar one and then hear every subsequent clue as agreement. Deliberately spend one attempt trying to disprove your leading candidate; if it survives that, it is probably right.

Speed comes last. Solve slowly and correctly for two weeks, narrating your reasoning aloud, and the speed arrives on its own as the eliminations become automatic. Solving fast by guessing builds nothing you can use in a session with other musicians.