Between June 6 and July 3, I shipped six daily puzzle games. A bird memory game, U2, Dash10, Plot30, ShiftLock, and 2D. Different mechanics, different audiences, different amounts of my weekends. They share exactly one design decision, and it is the only one I never debated: everyone gets the same puzzle, and it dies at midnight.

I did not invent this. Nobody I know invented this. But after building six games around the constraint, and DECIPHER before them, I have stopped thinking of it as a game mechanic. It is a social technology. The puzzle is the excuse. The constraint is the product.

Here is what the constraint actually does.

One puzzle a day converts a game into an appointment. When play is unlimited, a player binges, gets their fill, and leaves with nothing scheduled. More minutes, less attachment. When play is rationed, the game acquires a time slot in someone’s day, and time slots are sticky in a way that fun is not. U2 taught me this. It is a word splitting game a person finishes in two or three minutes, which by any engagement metric should be a weakness. Instead, 80 percent of its visitors are return visitors. Nobody returns daily to something because it is long. They return because it is theirs at a specific hour, with coffee, on a commute, in the gap before a meeting. Scarcity built the ritual. The game just showed up on time.

The sameness does separate work. Comparing results only means something if the test was identical, which is why every leaderboard I have built would be worthless without this constraint. A high score on a random puzzle is a fact about one person. A score on the shared puzzle is a position in a conversation. ShiftLock, a word ladder with a deduction layer, holds a similar 80 percent return rate, and I suspect the deduction is why. It produces defensible differences. Two players who both solved it can argue about the path they took, and the argument is only possible because the destination was fixed. Plot30 takes this to its logical end. It is a collaborative 30-word story, six or seven contributors a day, and there is no puzzle to solve at all. The shared constraint is the entire game. Small numbers, but those six people are building one artifact together, and you cannot do that with personalized content.

The third thing the constraint does is forgive. Midnight erases the board. Miss a day and you have lost one day, not accumulated a debt. There is no backlog of 14 unplayed puzzles glaring at you, no streak counter converting a habit into a hostage situation. Duolingo perfected the opposite model, guilt as retention, and it works, but it works the way a gym contract works. My games cannot afford to punish absence, so the erasure does something better. It makes coming back cheap. Every midnight is a clean slate, which means every morning is a fresh invitation rather than an accusation.

Now the part I would rather skip. This constraint is table stakes. Thousands of dailies exist, and every one of them resets at midnight and hands everyone the same puzzle. The constraint wins you an audition for a habit slot. It cannot keep the slot. My own six prove it. U2 and ShiftLock retained. Dash10, a speed round game, sees low turnout and I do not expect that to change. 2D, my version of the 24 game, is three days old, too early to call. The daily format did not save Dash10, and here is the uncomfortable mechanism: it accelerated the verdict. A game that is re-evaluated every 24 hours gets no honeymoon. Each midnight, the player asks whether tomorrow’s puzzle earned an appointment, and speed for its own sake apparently did not. The constraint is a multiplier. It made my good mechanics stickier and my weak one dead faster.

The constraint, it turns out, is infrastructure. It gives everything and guarantees nothing. Six games in, the pattern is plain: players forgive a modest puzzle if the appointment is good, and they cancel the appointment the moment the puzzle stops earning it. I thought I was building games. I was building appointments, and the games were the excuse to keep them.