Taking the losing out
Remove the moment where you lose and a puzzle usually stops being a game. These three kept their teeth by rehousing the two jobs that the ending was quietly doing.
Three games went up on this site today, and they are all built on the same edit: take a familiar puzzle genre and remove the moment where you lose.
That sounds like removing the game. It is not, and the interesting part is why — what a losing condition is actually doing, and what has to take its place when it goes.
What the loss was for
In a block-placement puzzle, the run ends when none of your pieces fit. In a sorting puzzle, you can pour yourself into a position with no legal move. In a merge game, the basket fills to the top. In all three, that moment is doing two jobs at once.
The first is scoring. The end of the run is when the number stops moving, so it is the number. Without an ending there is nothing to compare.
The second is consequence. It is what makes the placement you are considering right now matter. Remove it carelessly and every move becomes equivalent, which is the real failure mode of a cozy game: not that it is too easy, but that nothing you do is a decision.
So the rule for all three was: the ending goes, both of its jobs have to be rehoused.
Patchwork: the loss becomes a completion
In Patchwork you place felt patches on an eight by eight quilt frame and completed rows are stitched away. When none of your three patches fits anywhere, the classic version says game over.
Here the quilt is simply finished. It folds away with a sound, a counter goes up, and a fresh frame is laid out with a few patches already sewn in. Nothing is taken from you and there is no score to have been ruined.
Consequence survives because the thing you are trading is still real: every patch you place badly shortens the quilt you are working on. You are not avoiding a failure, you are deciding how much you can stitch before the cloth runs out. The difference between those two framings is entirely in the wording and completely changes how it feels to stop playing mid-quilt.
Scoring moves to the quilt count, which only ever goes up and does not care how good any individual quilt was. That is a weaker competitive hook and a much better reason to open the tab again.
Wool Jars: the loss was never necessary
Wool Jars is a sorting puzzle — pour wool between jars until each holds a single colour. This genre does not really have a loss so much as a dead end: you can pour yourself into a position with no legal move and have to restart.
Unlimited undo removes it entirely, and costs nothing, because in a puzzle with no timer the only thing undo takes away is the tedium of replaying moves you already knew were right.
The harder problem in this genre is a board that cannot be solved at all. Fill the jars with a random permutation and a meaningful fraction of boards have no solution — and a player cannot tell an unsolvable board from a hard one, so the game silently becomes unfair. The fix is to generate backwards: start from a solved arrangement and pour wool around at random. The shuffle is a solution played in reverse, so one always exists.
One more detail matters more than it looks. The generator originally left the empty jars as the last entries in the array, which put them in the same corner of the screen on every board. Every puzzle opened with the same move and looked like the same puzzle. Shuffling the jar positions after generation fixed a variety problem that no amount of extra colours would have.
Yarn Drop: the loss becomes a cost
Yarn Drop drops yarn into a basket, and touching balls of the same size wind together into one bigger ball. The genre ends when the basket fills.
Here, filling the basket means the cat lifts the top row away and play continues. A bad stretch costs you height and position — which is genuinely expensive, because the balls that get removed are the ones you were building on — but it does not cost you the session.
Consequence is entirely intact: a badly placed large ball is dead weight in a corner for a long time. What is gone is only the part where you have to start again from nothing.
The thing that nearly broke it
In Yarn Drop the chain of merges runs only from the ball you just dropped. Scanning the whole board for equal neighbours after every move is easier to write and it destroys the game — one tap tidies the entire basket, and choosing a column stops meaning anything. That is the cozy failure mode in its purest form: the game becomes pleasant to watch and pointless to play.
It also produced the best bug of the batch. When a group merges, the cells above it fall, and the merged ball itself can fall with them. The first version checked the next link of the chain at the coordinates where the merge happened, which by then was an empty square — so every cascade stopped after exactly one step and the game felt inexplicably flat. The fix is that the gravity routine takes the cell it is tracking and returns where that cell ended up.
What the automated checks said about all this
Every game here has to pass a play gate before it can be published, and that gate assumes a
game ends when you stop playing. All three of these declare session: 'cozy', which
inverts that rule — there is a whole post about
how that mechanism came to exist. What stayed strict is the requirement that something
advances while input arrives, which is the check that actually matters for a game with no
failure state.
That check earned its keep immediately. Wool Jars failed it: the automated tester tapped for eight seconds and never once made a legal pour. Two real problems were behind it — the empty jars always sitting in the same corner, and hit-testing that used rectangles so that taps landing between two jars did nothing at all. A tap that produces no response is the one genuinely frustrating event possible in a game with no timer, and it took a robot to notice.