40ants

40ants / microgpt

Public

A Common Lisp port of @karpathy microgpt.py — the most atomicimplementation of a GPT with a hand-written scalar autograd engine.

10
0
100% credibility
Found Mar 06, 2026 at 10 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Common Lisp
AI Summary

An educational Common Lisp program that trains a minimal AI model on a list of names to generate new invented names without using any specialized machine learning tools.

How It Works

1
🔍 Discover microgpt

You hear about a simple program that lets AI learn from real names and invent new ones, just like magic.

2
📱 Visit the guide

Head to the project's page where friendly instructions show exactly what to do next.

3
🛠️ Set it up easily

Follow one quick step to make the program ready on your computer.

4
🚀 Start the training

Run the program and watch it grab a list of names and begin learning patterns.

5
🧠 See the AI learn

Numbers on screen improve step by step, showing the AI getting smarter at predicting names.

6
✨ Generate new names

The program creates a bunch of fresh, made-up names that sound real and fun.

🎉 Enjoy your results

You now understand how a tiny AI brain works and can tweak it for more creative names.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 10 to 10 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is microgpt?

Microgpt ports Karpathy's minimal GPT to Common Lisp, downloading name datasets, training a tiny model with Adam optimization, and generating plausible names via CLI or REPL calls. Run `ros install 40ants/microgpt` then `microgpt --steps 1000 --samples 20 --temperature 0.5` for instant results, or load via Quicklisp for `(microgpt:run)`. It's pure github common lisp, no ML frameworks, solving the itch for dependency-free AI experiments akin to common lisp tutorial projects.

Why is it gaining traction?

This beats the Python original by 10x speed on identical runs, leveraging Common Lisp's raw performance without libraries—hook for fans debating common lisp vs clojure or common lisp vs scheme. CLI flags like `--input` for custom data and seeding for reproducibility make it dead simple, echoing common lisp cookbook recipes. Educational vibe draws from common lisp hyperspec and common lisp a gentle introduction to symbolic computation, minus bloat.

Who should use this?

Common Lisp devs prototyping microgpt ai, teachers demoing autograd via common lisp loop and common lisp format in classes, or Roswell users in common github actions workflows. Skip if chasing micro gpt coin price hype; ideal for common github usernames exploring lisp ML.

Verdict

Low 10 stars and 1.0% credibility score flag its immaturity—LLM-generated with sparse tests, so inspect common github labels and licenses before adopting. Worth a spin for fast CL AI baselines; clone with common github commands, tweak on common github branch names, and contribute common github tags.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.