- Number three on the brain-game shortlist (docs/brain-game-roadmap.md)
- You must predict before playing. Predicting afterwards is reporting, not thinking
- Two rules with different invariants: the jar hides its answer in the parity of the whites, the tokens in the parity of the total
- The big levels are past playing out by hand, which is the point: at that size the invariant is the only way through
- Both invariants brute-force verified: 25,800 exhaustive jar play-outs and 400 random token runs, zero violations
- A wrong call always gets the reasoning explained, and nothing is tallied against you
- A third rule: the chessboard-and-dominoes one, where the invariant is square colourplanned
✓ = done · ☐ = planned. Every game is built one step at a time like this.
