karpathy/scriptsbots — explained in plain English
Analysis updated 2026-08-15 · repo last pushed 2012-01-02
Run the simulation and watch creatures evolve survival strategies over thousands of generations.
Use it as a visual classroom demo to show how natural selection works.
Experiment with mutation rates and starting conditions to see how different ecosystems emerge.
| karpathy/scriptsbots | sunjaycy/goldeneye-recomp | peng-zhihui/grblhost | |
|---|---|---|---|
| Stars | 183 | 183 | 187 |
| Language | C++ | C++ | C++ |
| Last pushed | 2012-01-02 | — | 2021-05-26 |
| Maintenance | Dormant | — | Dormant |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 3/5 | 3/5 | 3/5 |
| Audience | general | general | vibe coder |
Figures from each repo's GitHub metadata at analysis time.
Requires C++ build tools and OpenGL setup, README provides step-by-step instructions for Linux and Windows.
ScriptBots is a digital petri dish where virtual creatures live, hunt, eat, and evolve entirely on their own. Created by Andrej Karpathy, it's a simulation of predator-prey dynamics, imagine watching a tiny ecosystem unfold on your screen where little bots wander around, some hunting others, some fleeing, all competing to survive and pass their traits to the next generation. At a high level, the simulation runs an evolutionary process. The bots have basic behaviors driven by internal "brains" that can mutate over generations. Bots that are better at finding food or avoiding predators live longer and reproduce, passing their characteristics to offspring. Bots that fail die off. Over time, you can watch survival strategies emerge and shift as the population adapts, all without anyone programming those strategies directly. This would appeal to anyone curious about evolution, artificial life, or how complex behaviors can arise from simple rules. A hobbyist might run it just to see what kinds of creatures emerge over thousands of generations. A student or educator could use it as a visual demonstration of natural selection. A founder or PM interested in AI might find it a tangible way to grasp how evolutionary pressure shapes behavior, the same underlying concepts that appear in some optimization techniques. The project is written in C++ and uses OpenGL for its visuals, meaning you see the ecosystem play out in real time on screen. Building it requires a few developer tools, and the README provides step-by-step instructions for Linux and Windows. It doesn't go into much detail beyond setup and controls, so you're largely on your own to experiment once it's running. That hands-off approach fits the spirit of the project: set it up, press play, and see what nature comes up with.
A virtual petri dish where digital creatures hunt, eat, reproduce, and evolve on their own. Watch natural selection and survival strategies emerge in real time without anyone programming them.
Mainly C++. The stack also includes C++, OpenGL.
Dormant — no commits in 2+ years (last push 2012-01-02).
No license is specified in the repository, so default copyright restrictions apply and you should contact the author before using or distributing the code.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly general.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.