Games with Laya

A project where Laya-MLX plays the three-player Chinese card game Dou Dizhu, making each play from candidate options.

Combines screenshot perception, Laya decisions on Core ML and Qwen planning to play, click loot and review its own evidence, entirely locally.

Daniel Tremer's autonomous agent for Hordes.io: Qwen3.5-4B on MLX plans, Laya on Core ML decides, Apple Vision reads the screen. All on a 48 GB M3 Max, no cloud.

Trains the 421M Laya ModernBERT model to choose Balatro actions from candidate lists, using simulator self-play and a real-game test set.

A pixel space shooter whose ship is controlled by Laya through a local /v1/systemone API, forked from a Jev version of the same game.

Uses Laya's typed decisions to work through murdle.com's daily murder-mystery logic puzzles.

Native Windows demo with live probabilities, executed moves, inference timing and planner-intervention counts. A deterministic search proposes routes, Laya classifies them.

Backend-independent Unity client for local Laya servers and TypeSafe-compatible endpoints. Game state in, Choice/Score/Noul out; transport-first, no embedded weights.

A survival arena game where every enemy's next move is a live typed question put to Jev or Laya, with probabilities coming back instead of text. Lets you watch the two decision models play head to head.

Soybelli connected Pac-Man to laya-mlx on Apple Silicon. The live view shows action probabilities, latency, tactical planning and safety interventions, with 0 output tokens.

Side-by-side Snake: Laya local off a 1 GB footprint made 86.4 decisions a second (1,281 moves); Jev 1.13.0 over the API managed 3.2 (47 moves).

From the Prompt Engineer 48 video. Both models get the same input and typed questions; only the model differs. Laya needs no key.

Local 15x15 Gomoku on Apple Silicon where each AI move is a single Laya choice among up to six candidate moves, with probabilities shown in the browser. The model was never trained on Gomoku.

Self-hosted console that runs VirusTotal lookups through System One models and evaluates them against each other on a labelled dataset. Prompts and criteria are versioned JSON.

The game enumerates placements with their holes, heights and clears; Laya picks one from text. The teacher is a board-evaluation formula, not Jev. Walkthrough video in Japanese.

Brain function collapse's live site where Laya plays Tetris, Flappy and a lane runner on a laptop, about 200x faster than a hosted LLM. Backed by the laya-playground repo.

MLX port of Laya with performance work. Under 1 GB of memory on-device; the demo plays Snake at 60 decisions per second on an M3 Max.

Abhimanyu walks the Mahabharata ring formation to the centre. Each step goes to local Laya as typed questions; unsure or unplayable answers are marked red and corrected.