Games · AI · Evolution simulation game / personal project

Gene Hive

A hex-based insect colony simulation you watch rather than command: autonomous workers forage, build, breed and defend the nest, and every insect's behaviour, body and small neural brain come from a genome that evolves across generations.

26 genes per insect

Godot 4.7 C# Genetic algorithms Neural networks

Gene Hive following one insect near the Nectar Vats and Hive Heart, with the Evolution panel (population, generations, genetic variety, lineages, how much the brain steers) on the left and the selected specimen's neural network from senses to goals on the right

Zoomed-out view of the colony on the hex map with the danger overlay, next to a selected insect's decision diagnostics, time spent per goal and genome sign

The hex world with grid lines at night and the settings panel open: reduced motion, world effects, hex grid lines, health bars, interface size, controls and audio volumes

Close-up of the nest: Nest Gate, Chitin Vault, Hive Heart, Resin Store, water and brood cells, with insects carrying resources between them

Stats panel after 248 generations showing population, genetic variety, predation pressure and how much the brain steers, beside a wide view of the colony

Role: Author, developer & owner
Platform: Windows (Godot)
Status: Playable foundation, in development
Workflow: AI-assisted, agent-driven

Overview

The player is an observer with a classic strategy-game camera over a hex world. The insects decide for themselves: they satisfy hunger, thirst, fatigue and social needs, gather nectar, water and resin for the colony, raise brood and fight off spiders. What drives those decisions is inherited, so a colony slowly becomes better at staying alive.

Genes you can see

Behaviour genes set how readily an insect forages, helps kin, shares food, raises the alarm or answers the colony’s needs
Morphology genes change the body itself: eyes, legs, abdomen, mouthparts, antennae, armour and defensive spines, each with a real effect on sensing, speed, carrying, disease resistance or combat
A small neural brain (11 inputs, 4 hidden, 11 outputs) steers decisions alongside authored scoring, and it learns during the insect’s lifetime
Painted bodies built from an artist-editable part rig show every inherited feature, with age visible too: freshly hatched workers look smaller and paler

A living world

Seasons, weather and a day/night cycle affect thirst, plant growth and harvest yields
Plants regrow, and corpses decay back into soil fertility
Ant-style pheromone trails, fungal spores that insects learn to avoid, and roaming spider predators that stalk, attack and retreat
Social behaviour emerges rather than being scripted: food sharing, grooming, alarm recruitment and collective defence
Two healthy adults with enough stored food produce a genetically crossed brood; immigrants keep the gene pool from running dry

Built to be measured

The simulation runs on a fixed 60 Hz deterministic clock with seeded random streams, so every run can be reproduced exactly and saved mid-flight. The same rules run headless for offline evolution: a genetic algorithm (with optional NSGA-II multi-objective selection) trains champion policies, and test-only tools hunt for exploits, measure throughput and compare runs across seed banks.