Releases·Tool Release
Show HN: A TypeScript PyTorch-style runtime that trains and runs inference in-browser on WebGPU
A developer launched a browser-based runtime that executes PyTorch-style code via WebGPU, with no installation, server, or account. The TypeScript-based tool supports both model training and inference and was built to let students run torch code in classes without setup.
1 engagement1 sourceWed, Sep 30, 2026, 04:15 AM
AI Overview
- Show HN author says the WebGPU TypeScript runtime trains models and runs inference with no install, server, or account, built for classroom use
Related
Releases·Tool Release
Tynx: train ONNX models with a PyTorch-shaped API in a <20 MB wheel
Tynx is a new Rust-based ONNX runtime that offers a PyTorch-shaped API for both inference and training, packaged in a wheel under 20 MB. It uses Burn/CubeCL and wgpu for GPU execution across OSes and GPUs without extra libraries, and can compile to the browser.
Tynx is a new Rust-based ONNX runtime that offers a PyTorch-shaped API for both inference and training, packaged in a wheel under 20 MB. It uses Burn/CubeCL and wgpu for GPU execution across OSes and GPUs without extra libraries, and can compile to the browser.
1 engagement1 sourcehackernewsWed, Aug 12, 2026, 05:31 AM
Releases·Tool Release
Software Mansion releases runntime, a TypeGPU-based framework for running AI models on the user's GPU in WebGPU environments
Software Mansion released ruNNtime, a framework for running AI models on the user's own GPU via WebGPU, built on TypeGPU. Kernels are written in TypeGPU and compiled to WGSL at runtime, so no WASM binary is needed and the same code runs in the browser, React Native, and Electron. It offers a 'zoo' of one-function-per-task models for speech, vision, and NLP, positioning itself as an alternative to transformers.js for on-device web inference.
Software Mansion released ruNNtime, a framework for running AI models on the user's own GPU via WebGPU, built on TypeGPU. Kernels are written in TypeGPU and compiled to WGSL at runtime, so no WASM binary is needed and the same code runs in the browser, React Native, and Electron. It offers a 'zoo' of one-function-per-task models for speech, vision, and NLP, positioning itself as an alternative to transformers.js for on-device web inference.
37 engagement1 sourcegithubMon, Sep 14, 2026, 01:28 PM
Products·launched
Client-side poker solver using custom WebGPU kernels generated by Codex
A poker solver that runs entirely in the browser via WebGPU, hosted for free. The developer used Codex to convert thousands of lines of PyTorch code into custom WebGPU kernels with parity tests, then optimized them over days. It solves a real problem for poker players and developers who need a free, client-side solver without server costs.
A poker solver that runs entirely in the browser via WebGPU, hosted for free. The developer used Codex to convert thousands of lines of PyTorch code into custom WebGPU kernels with parity tests, then optimized them over days. It solves a real problem for poker players and developers who need a free, client-side solver without server costs.
95 engagement1 sourcehackernewsThu, Jul 30, 2026, 03:38 PM
Releases·Tool Release
Hugging Face open-sources WebGPU kernel collection for fully local browser AI, plans upstreaming to Transformers.js and ONNX Runtime Web
Hugging Face released open-source WebGPU kernels claimed to be the fastest for local AI, covering 200+ common ML operations that run entirely in the browser. The team plans to upstream the optimizations into Transformers.js, ONNX Runtime Web, and LiteRT.js, making fast local inference more accessible to web developers.
Hugging Face released open-source WebGPU kernels claimed to be the fastest for local AI, covering 200+ common ML operations that run entirely in the browser. The team plans to upstream the optimizations into Transformers.js, ONNX Runtime Web, and LiteRT.js, making fast local inference more accessible to web developers.
624 engagement1 sourceredditWed, Sep 30, 2026, 04:02 PM
Products·building
Project Mosaic: Distributed AI compute in browser via WebGPU
Project Mosaic is a platform that runs distributed AI compute workloads directly in the browser using WebGPU, enabling consumer devices to participate in training or inference without dedicated GPU clusters. It uses quantization and sharding to keep memory usage low (~1.8GB per tab) and aggregates updates from nodes centrally. It aims to make AI compute more accessible and cost-effective by leveraging idle consumer hardware.
Project Mosaic is a platform that runs distributed AI compute workloads directly in the browser using WebGPU, enabling consumer devices to participate in training or inference without dedicated GPU clusters. It uses quantization and sharding to keep memory usage low (~1.8GB per tab) and aggregates updates from nodes centrally. It aims to make AI compute more accessible and cost-effective by leveraging idle consumer hardware.
4 engagement1 sourceredditWed, Aug 26, 2026, 12:47 AM