Back to browse
Watch a neural net learn to play Snake

Watch a neural net learn to play Snake

by c1b·May 14, 2026·205 points·47 comments

AI Analysis

●●●BangerWizardryRabbit Hole

Runs PPO training entirely in-browser via TinyJit WebGPU kernels.

Strengths
  • TinyJit compilation pipeline enables complex ML training client-side.
  • Real-time visualization of policy, value, and entropy metrics.
  • Zero server dependency for the actual reinforcement learning loop.
Weaknesses
  • Strictly requires WebGPU adapter, excluding many older devices.
  • Niche educational demo rather than a production tool.
Category
Target Audience

ML engineers and WebGPU enthusiasts

Similar To

TensorFlow.js · WebLLM

Post Description

In browser PPO training demo, made possible by tinygrad: TinyJit -> WebGPU kernels.

Requires WebGPU.

Similar Projects

GamingMid

Play Bang Bang against a neural network

You play a minimalist web game while a neural net trains in real time, so the opponent starts weak and adapts to your playstyle. The landing page is extremely spare — the demo is quick to load and oddly charming, but there’s little signposting about the model, implementation details, or source code. Fun as a curiosity or teaching toy, but not novel enough to wow ML folks who’ve seen in-browser learning demos before.

Niche GemCozy
atum47
104mo ago