A screener built to resist the usual false positives — randomised octaves and timbres, a memory-wiping scramble between trials, latency capture, and no feedback while it's running. Then a training track that starts from one landmark note and expands.
Run this cold. Not after training, not right after playing your keyboard, not while a track is stuck in your head. It gives you no feedback until the end — feedback during a test is training, and would contaminate the measurement.
listen
One note learned properly beats twelve learned vaguely. You start by deciding only whether a tone is C or not, then add notes that are maximally far apart before filling in the gaps. Feedback is immediate here, and the scramble still runs between trials so you can't chain one answer off the last.
press start
No completed screens yet. Run the baseline first.
Nothing persists when this page closes. Export before you finish and paste it back next session so the re-screen has a baseline to compare against.