Same start. Own pieces.
One seed builds the board and the first hand. After that, you and the AI draw different pieces, so nobody can copy a move.
Same start. Different pieces. You and a neural network, side by side. Whoever plays better wins.
Every match starts identical. Same board. Same first hand. Then the pieces split, so there is nothing to copy. Just two players, one human and one neural network, and a score that settles it.
One seed builds the board and the first hand. After that, you and the AI draw different pieces, so nobody can copy a move.
The model runs locally in a background thread, so the board never stutters while it thinks.
Static files. Open it anywhere.
The AI imagines whole hands before it commits to a move.
Every board has a link. Send it, and they play your exact game.
I build things I wish existed. This puzzle started as a Python experiment. Then I taught a neural network to play it, squeezed it down to 711 KB, and put it right here in your browser.
No servers, no accounts, no tracking. Just a game, a machine, and a score.