Look, then put it back
A pattern appears, then goes. Reproduce which cells were marked, or the order they came in, or complete a half-pattern across a stated mirror line. No score is kept and nothing is claimed about what any of it measures.
Nothing is recorded or sent anywhere
Why symmetrical patterns are thrown away
Not tidiness — it is the only thing that keeps the difficulty label honest.
Nine cells arranged symmetrically are far easier to hold than nine scattered ones, because you remember four positions and a rule rather than nine positions. So a generator that produces a symmetric arrangement by accident has quietly handed out an easy round with a hard label, and there is nothing on screen to tell you that is what happened.
Every pattern is therefore checked against all four mirror lines and both quarter-turns, and regenerated if it lands on one. It is not a rare event: six cells dropped at random on a 5 × 5 come out symmetric in about one draw in seventy — measured, not estimated, and the test suite measures it rather than taking the claim on trust.
Patterns bunched into a corner are rejected for the same reason. Four cells in a 2 × 2 block are one shape to remember, not four positions.
Three numbers instead of a score
Cells found, cells missed, and cells you added that were not there. A single percentage hides the difference between someone who remembered four of six and someone who remembered all six and added three more — those are not the same performance, and rolling them together is the sort of tidiness that loses the only interesting information.
In order mode there is a fourth thing worth knowing: whether you had the right cells in the wrong order. That is a genuinely different outcome from having the wrong cells, and it is reported as such.
Nothing is stored. Close the page and the round is gone.
What this is, and is not
It is a pattern game. It is not a test, not an assessment, and not training — no claim is made that playing it improves anything, because the evidence for transfer from tasks like this to anything you would actually want to be better at is weak and much-disputed. Results here describe the round you just played and nothing beyond it.
There is no streak, no daily target and no notification. If you stop, nothing happens.
Ways people actually do it
Most people start by trying to hold the positions and switch, somewhere around six cells, to holding a shape — reading the marks as a letter, a constellation, a route. That is why symmetric patterns are filtered out: they hand you a shape for free, and the count stops describing the work.
In order mode the useful trick is different. Positions are held in one way and sequences in another, so people who are good at the first are often no better than average at the second. Chunking helps — three pairs rather than six steps.
Untimed study mode exists because rushing is not the interesting part. Look for as long as you like, press hide when you are ready, and the puzzle is the same one.
Reading it without colour
Every state carries a shape as well as a tint: marked cells have a dot, correct ones a tick, missed ones a dashed outline, and cells added in error a cross. The mirror line in symmetry mode is drawn as an actual line rather than implied by shading, and the board is a proper grid for screen readers, which announce each cell's row and column and its state.
Related
Nonograms for a picture you deduce rather than remember, or the puzzle hub for everything else.
Questions people actually ask
Does this train or measure my memory?
No, and nothing here suggests it does. The evidence that tasks like this transfer to anything you would actually want to be better at is weak and much-disputed, so no claim is made. Results describe the round you just played and nothing beyond it.
Why does it reject symmetrical patterns?
Because a symmetric arrangement is remembered as half the cells plus a rule, so it is much easier than its count suggests — and the difficulty label is built on the count. Every pattern is checked against four mirror lines and both quarter-turns. It matters more than it sounds: six cells dropped at random on a 5 × 5 come out symmetric about one draw in seventy.
Why three numbers instead of a score?
Because remembering four of six is not the same performance as remembering all six and adding three more, and a single percentage hides exactly that. You get cells found, cells missed, and cells added that were not there — plus, in order mode, whether you had the right cells in the wrong order.
Can I look for longer?
Yes. Any setting can be given six or ten seconds, or made untimed so you hide the pattern yourself when ready. The puzzle is identical either way; only the time to look changes.
Does anything flash?
No. Cells fade in and out over about a fifth of a second, nothing blinks, and in order mode each cell stays lit for most of its slot. Reduced-motion settings turn the transitions off entirely.
Is a streak or a history kept?
No. There is no account, no streak, no daily target and no notification, and nothing about a round is stored. Close the page and it is gone.