Exhibit 05Adversarial Search
Tic-Tac-Toe
Play Jev yourself, or watch it face a perfect minimax player. Tic-tac-toe is a solved game: perfect play is always a draw, so every loss is a mistake someone can point to.
Partly in scopeBoard-only play is the chess-like planning TypeSafe says suits reasoning models; with tactics from code it is a snap choice. TypeSafe homepage FAQ: “Some tasks requiring extended reasoning, such as complex mathematics or chess-like planning, may be better suited to large reasoning models.”
World
→State
→Decision
→Action
→New state
Jev playsJev is
Context given to Jev
The board, which mark Jev plays, and the empty squares.
Now
Your move · you are X
JevYou
vs You · Board only
0
Jev
0
Draw
0
You
0 gamesblunders 0 / 0 moves
Jev decision
idleYou are X and move first. Click a square.
top left—
top center—
top right—
middle left—
Other (5)—
Lesson · Adversarial
My move ↔ your best reply
Tic-tac-toe is small enough that minimax solves it completely: from the empty board, perfect play is a draw. That makes every Jev move gradable. A blunder is any move that lowers the perfect-play outcome: a drawn position turned into a lost one, or a won position thrown away. The coach says so. The context levels mirror Missionaries: the board alone, then facts code can check one move ahead for each side (wins, blocks, the opponent's immediate replies, double threats), then the perfect-play result itself. The question is the same as before: how much of good play is judgement, and how much is the lookahead we hand it?