What is anti-human chess?
For 30 years we've talked about anti-computer chess: closed positions, no tactics, make the engine drift. Anti-human chess is the mirror image. The bot doesn't look for the objectively best move. It looks for the move that makes you most likely to go wrong.
Engines don't get tired, don't get anxious and don't have a working memory limit. Humans do. So an anti-human bot plays to those limits:
Out of the book early. Sound sidelines instead of main lines, so your preparation is useless by move 8.
Keep everything on the board. No trades, no simplification. Pieces stay, pawn tension stays. Every move you have 6 plausible candidates instead of 2.
Never release the tension. Humans resolve tension because it's uncomfortable. The bot leaves it there for 20 moves and lets you crack first.
Two problems at once. One weakness, you defend. Two weaknesses on opposite flanks, you pick the wrong one.
Quiet moves. Not a sacrifice you can calculate to the end, but a prophylactic rook move with no obvious threat. What does it want? Now you're spending clock.
Convert cleanly. Once you've made the mistake, it trades into a won endgame and the fog lifts.
We tested it. At 2000 Elo, The Anti-Human scores 44% against Stockfish, the same as the plain Camille engine. Against that same engine with Camille's human-error model switched on, it scores 71%. Same strength, same thinking budget. It just makes the opponent go wrong more often. It's not playing the board, it's playing you.
In Camille, The Anti-Human is one of the bundled bots. It's built from a four-phase plan (sideline surprise fog of war the squeeze clinical conversion) and you can open it in the bot editor to see exactly how it's wired, or tune it down to your level.
Try it and tell me how many moves it took before you felt the pressure.