lgit

lgit is an interactive demo version control tool where code and images are stored inside a neural network.

lgit is an interactive demo of version control where the storage mechanism is a neural network instead of traditional bytes. Each commit runs a training process where the network learns to map positional encodings to your data, whether byte by byte for code or pixel by pixel for images. Checking out data performs inference, reconstructing files from the network weights with varying degrees of accuracy. Users can commit code, train on images, adjust hidden layer sizes, and observe unique types of neural corruption that differ from traditional file corruption. The system uses Fourier positional encoding, ReLU hidden layers, and an Adam optimizer entirely within the browser. It is designed for developers and curious users who want to see how a neural network remembers and reconstructs committed data.

lgit at a glance

Best for: Software engineer
Runs on: Web
Pricing: Free
Released: October 2026
Category: AI coding tool
Website: nicobrenner.github.io

Key features

Commit code into a neural network
Train on images with real-time reconstruction
Adjust network size with hidden layer parameters
Use Fourier positional encoding and ReLU hidden layers
Optimize data mapping with an Adam optimizer
Observe neural corruption during checkouts

Use it for

Commit code into a neural network to see what it remembers
Train on images and watch the reconstruction in real time
Test neural corruption effects on source code

Alternatives to lgit

Git AI — It moves work to the server and handles concurrent tasks without standard Git bottlenecks.
jevgrep — Reduces coding agent costs by providing targeted, relevant repository context.
FinderGit — Review and manage all your Git repositories directly in a native macOS file browser.
GitAgent — Manage AI agents with familiar git workflows like branching, diffing, and reverting.
AI Memory — Shared memory for coding agents that works across machines and team members.
TheGitAI — Operates active browser sessions without requiring APIs or credential sharing.
Entire CLI — Mirrors Git repositories globally to increase cloning speeds for AI agents.
OpenCode — Works with 75+ model providers, local models, and existing subscriptions
LangChain4j — Integrates LLMs and vector stores directly into Java applications.
uithub — Provides immediate full repository context for LLMs by changing one URL letter.
Weft — It enables safe parallel coding agent work without file locks.
guides.show — Explains code diffs in order of importance through portable, encrypted links

All lgit alternatives