Builds made with Laya

本地语境拼音输入法：改进 lime 候选召回，支持 Laya MLX 实验重排；Apple Silicon + 鼠须管。

Snake where every move is decided by a fine-tuned copy of the Laya decision model

Modern Sudoku game solved in real-time by Laya System 1 decision model (convaiinnovations/laya) with ~35ms GPU latency.

Offline Laya Snake demo for SpaceMIT K3: reproducible Bianbu RISC-V setup, ONNX AI Core adaptation, and tests

Two System One models fight a real Doom deathmatch. Laya (322M, open weights, local MLX) vs Jev (TypeSafe hosted). Same state, same typed questions, same shield — only the model differs.

I made Laya (Jev-like model) play Breakout, Every paddle move is a live decision from an open 322M System One model (convaiinnovations/laya) running locally on Windows laptop, ~70 decisions/sec on an RTX 5080, zero cloud, $0.
Code 👉

Route each ADK request to the right Gemini tier with Laya on a CPU-only container (Cloud Run), plus an eval scored by cost per successful task

I had a stupid idea:

What if, instead of asking an #AI to talk, I made it fight me? 🥊

So I built World Summit Tournament.

21 fictional world-leader-inspired fighters, signature moves, combos, replays — and the opponent can use Laya, a small open-weight decision model in the…

1,000 synthetic emails. Six folders. One local 322M Laya model running on a MacBook Air M3.

⚡ 28.6s without screen recording
🎬 97.3s while recording
🎯 65.1% accuracy against reference labels
💾 0 MB swap

Everything runs locally. No cloud inference.

And I’m showing the…

Confidence-gated GitHub issue triage powered by a fine-tuned Laya decision model

Multilingual support ticket triage with a local System 1 decision model: hierarchical intent routing, urgency/frustration/churn signals, confidence-based escalation

Ultima Online duel bot driven by System One decision models (Laya local, Jev cloud), with a live duel monitor

Fully local AI coding assistant for a CPU-only laptop: Laya + LangGraph + Qwen 3 8B (Ollama) + MCP sandbox, with a kid-friendly UI

Local voice assistant for Linux/Hyprland: hotkey sessions with faster-whisper, a Laya typed-decision router, Ollama and Piper TTS. Runs fully offline.

Ghosthand (powered by LAYA) ⚡ Sub-100ms local voice assistant & computer use agent for macOS on Apple Silicon. Physical mouse gliding, click ripples, 12 window tiling modes, and frosted HUD.

Laya MLX typed decisions vs Dijkstra on seeded weighted mazes, side-by-side replay on Apple Silicon

Laya (local, on-device MLX) vs Jev (cloud API) controlling identical cities in a real-time AI traffic simulation

An open-weights System 1 decision model (LAYA, 421M) playing Doom in a real-time control loop on Apple Silicon. After TypeSafe's Jev x Doom demo.

Shrink Convai's Laya typed-decision model to run on single-board computers: BPE-merge-closed vocabulary pruning plus block-wise int8, evaluated on human-labelled robot commands with majority-class baselines.

A Rich terminal Tetris environment for local Laya MLX play on Apple Silicon.

ScreenQuest: an autonomous screenshot-to-action AI agent for https://t.co/RHiWnwGPEf, running locally on an M3 Max (48 GB).

Qwen3.5-4B / MLX + Laya / Core ML + Apple Vision OCR. No cloud inference.

Code:

Built laya-drift to visualise and monitor agentic-drift today.

It continuously measures how far an OpenCode session has drifted from its original plan using Laya + Jensen-Shannon divergence