Hobbyist's virtual worms with neural networks reproduce classic optimal foraging research

neuroecology · x · 2026-11-21

A developer built virtual animals in their spare time: digital worms with individual neural networks that must gather enough protein, carbs and lipids from different foods, receive neuromodulator signals from organs, evade predators with their own neural nets, and interact with each other.

Trained only on eating, egg-laying and predator avoidance, the simulation reproduces classic optimal foraging (ARS) findings — the worms learn to forage and lay eggs where it pays off long-term.

Related event: Developer Builds Neural Network-Driven Virtual Worm Ecosystem(2 posts)→

Original post →

More from Fun

Fun channel →