Pinned Loading

1. zero-tvm zero-tvm Public

   Phi-3-mini decoding in a browser on 10 hand-written WGSL kernels (27 files, 3,078 LOC). 228 dispatches/token at ~40 tok/s on M2 Pro — 22% behind WebLLM's TVM-autotuned 85 kernels on identical weights.

   WGSL 2 2

2. webgpu-kernel-fusion webgpu-kernel-fusion Public

   Research umbrella for GPU kernel fusion — eliminating per-dispatch overhead in sequential GPU workloads, from RL to transformer inference.

   JavaScript 3

3. gpubench gpubench Public

   Real WebGPU compute benchmarks in the browser. 5 compute workloads + fused transformer decoding across 4 kernel variants — 159–720× over PyTorch on the same GPU.

   JavaScript

4. webgpu-transformer-fusion webgpu-transformer-fusion Public

   Single-kernel fusion for autoregressive transformer decoding via WebGPU

   JavaScript 1

5. webgpu-p2p-evolution webgpu-p2p-evolution Public

   Browser-to-browser distributed evolution. WebGPU compute + WebRTC P2P. 4 demos, 113-line relay. No install, no CUDA.

   JavaScript

6. safenpm safenpm Public

   Drop-in npm install replacement that sandboxes postinstall scripts, blocks network access, detects typosquats, and catches supply-chain attacks before they execute.

   HTML 1