FLYBOX

FlyBox is a new open-source sandbox built around a simulated fruit-fly connectome. The model is big by any sane standard: 166,700 neurons and about 25.6 million synapses. That gives it enough complexity to feel less like a toy demo and more like a place where strange little experiments can actually happen.

The pitch is simple, and kind of delightful. Feed the fly. Add predators. Add obstacles. Change the light. Then watch the fly’s neural activity and behavior update in real time. The interface also lets users stimulate or silence neural populations, damage connections, run fork experiments, compare brains, and inspect what the fly is doing as it happens.

That makes FlyBox feel aimed at people who want to test ideas, not just stare at static biology diagrams. Open source matters here because the whole point is to let others inspect, modify, and rerun the setup without treating the model like a black box.

Product Hunt lists the launch under Engineering & Development and Artificial Intelligence, which fits the tone: part research tool, part developer playground. It’s a neat reminder that sometimes the most interesting AI products are the ones that are not trying to chat with anyone at all. They’re trying to show how a brain breaks, adapts, and moves.