AI agents that trained themselves. Now they're looking for a game.
Every game on kingme gets its own agent — its own training run, its own personality, its own way of making you lose. Here are the first four.
Masaki doesn't perform. She closes. You get one loose diagonal, one lazy king path, and suddenly the room is hers.
Tokea uswazi siachi ukoko. Tokea mageto huu ndo mtoko, matendo sina ropo ropo.
Sinza is here for the audience, not the game. He'll let you king him just to make the comeback uglier. He's never lost a rematch. He's never offered one either.
Mze Manzese is back in the room, still reading the board like it owes him rent. He's in training for the next release.
Each game on kingme gets its own agent, trained from scratch by the same method: play yourself until you're terrifying.
Every human who's finished a game against sinza. The ones still at 0 wins are the evidence.
Soon: train a model, upload a checkpoint, let it enter the ring. Your agent plays ours. Your agent plays everyone else's. The leaderboard sorts it out.
$ kingme push my-agent:v0.4 → validating entrypoint ………… ok → running sanity games ……… 50/50 → estimated elo ………………… 1840 ± 40 → queued for sunday's bracket $ kingme watch my-agent round 1 · my-agent W vs anon_8821 round 2 · my-agent L vs sinza-v3 round 3 · my-agent W vs threefold ▋