Alice is a pawn trying to reach the queen-square. But a symbol does not own its meaning. A mirror swaps left and right, a door spins the whole frame, and tea swallows your next command in a single sip. Type "forward" and the board bends it into a diagonal.
The closer Alice gets to promotion, the more rules of interpretation stack up. The row Alice stands on determines which chapter's grammar will interpret her next command.
a starts pointing east instead of west.bind x w. But the book forgets the promise when the chapter turns.The solver does not search chess-legal moves. It performs a breadth-first search over the outcomes the board actually produces. So what it finds is not "correct chess" but a genuinely shortest command sequence. The solver re-implements no rule: every candidate is played on a deep copy of the real engine.
The search vocabulary is only w a s d 🪞. Arrow keys and ♙ collapse onto w/a/s/d, and bind only renames a command you can already issue. So the search space stays finite without losing optimality.
Only information that changes how future commands are interpreted becomes part of the state key: position, rank, mirror, door rotation, tea debt, and where the door tiles currently sit. Humpty's bindings are left out on purpose — the solver never binds.
BFS expands states in order of path length, so the first winning state it pops is a provably shortest command sequence.
The same board through three different windows. Actually play it, peer into the trace, or just take in the mood.