The Room That Played Like a Worksheet

The most important entry in the whole log is four lines long, and it's a contradiction:

textloop-log · ace_of_clubs · room 1  (a Latin-square puzzle)

  AI judge:                          A−
  Human, playing the real room:      STUCK
      "two 4×4 grids, both look like a valid Sudoku solution —
       what's the difference?"
  Honest round-1 grade:              F  —  rule stated on the wall

The studio's own judge passed this room with an A−. Then a person sat down and played it and could not solve it. I went looking for which of them was wrong. The answer is neither, and that's the whole point.

Two ways to be wrong at once

The room was a Latin square — a Sudoku, essentially — and the studio had done the obvious thing: it printed the rule on the wall and put the grid underneath. The judge, which already knew the mechanic, read the wall, confirmed the puzzle was solvable, and passed it. The human, who did not already know the mechanic, read the same wall and felt like they'd been handed a homework sheet with the answer box printed at the bottom. When the twist turned out to be a hidden diagonal constraint, they had no way to know where to even look.

The record names the failure precisely — "rule stated on wall (frame-leak)" — and then names the fix that is not the fix: you can't solve it by hiding the rule harder, because then the player is stuck for the opposite reason. There are two failure modes, not one. Too loud is a worksheet. Too quiet is a wall you bounce off.

The sentence that became the studio's spine

Out of that one stuck playtest came the line that now headlines the studio's design canon, written a few days later:

Puzzles are things they encounter and manipulate, not problems presented to them with instructions and an answer box. The difference between a game and courseware is entirely here.

That's the whole studio reorienting around a single word a tester didn't even say out loud. And it forked immediately into two problems the rest of this record is about. One was physical: if the puzzle can't be a grid printed on a wall, it has to be built out of objects — which meant the studio suddenly needed rooms that were real, measured, three-dimensional places. The other was conceptual: how do you put a rule into a room without ever stating it?

The studio chased both at once, in the same 72 hours. The next file is the physical one — the day it stopped drawing rooms and started measuring them.