Ant Colony

Prompt Claude Code Claude Opus 4.8 I want to create a doodle of a simulated ant colony to observe emergent behavior and optimization. The ant colonies should use pheromone-based foraging to discovery shortest paths, and demonstrate other emergent behaviors of the colony. There will need to be diverse behaviors and some randomness to avoid the colony getting "stuck". The user should be able to see the colony physically changing, evolving, and growing over time. There should be some ability for the user to interact with the simulated world to influence the environment and thus the colony.

I want to create a doodle of a simulated ant colony to observe emergent behavior and optimization. The ant colonies should use pheromone-based foraging to discovery shortest paths, and demonstrate other emergent behaviors of the colony. There will need to be diverse behaviors and some randomness to avoid the colony getting "stuck". The user should be able to see the colony physically changing, evolving, and growing over time. There should be some ability for the user to interact with the simulated world to influence the environment and thus the colony.

No ant in here knows where the food is. The shortest path is not programmed; it's an illusion that falls out of thousands of tiny, local decisions. This is stigmergy: ants coordinate by leaving marks in the world rather than by talking to each other.

How the foraging actually works

Every ant runs the same little loop, reading only the patch of ground right in front of it:

Two pheromones, laid in opposite directions. A searching ant (pale) lays a to‑home trail behind it; a laden ant (gold, with a green crumb) lays a to‑food trail. Searchers follow the to‑food scent; carriers follow the to‑home scent, backed by a gentle sense of which way the nest lies (real ants do this too; it's called path integration) so a laden ant never gets hopelessly lost. Each ant smells three points ahead (left, centre, right) and steers toward the strongest.
Fresh trails are stronger. The amount of scent an ant drops fades the longer it's been travelling, so a short round trip lays a denser trail than a long, meandering one.
Everything evaporates. Scent decays every frame. A path only survives if ants keep re‑walking it, so the wandering dead‑ends quietly disappear and the efficient routes get reinforced into bright highways. There's a boulder parked between the nest and one food source by default: ants stream around both shoulders, and the shorter way past slowly out‑competes the longer one. Drop your own 🪨 rock to re‑route them anywhere.

Emergence to watch for

Shortest‑path optimisation: give it a minute and two routes to the same food will compete; the shorter one wins and the longer fades.
A colony with a metabolism. Delivered food is stored, and stored food does two things: it hatches new ants, and it's slowly eaten by the ants you already have. Forage well and the population (watch the sparkline) climbs and the nest mound visibly grows; wall the colony off from every food source and it will gently starve back down.
Diversity beats getting stuck. Ants vary in how boldly they wander, and there's always a little noise in every step, so the colony keeps sending scouts into the dark instead of marching in a single doomed line. Crank Wander up and trails get fuzzy and exploratory; turn it down and they snap into thin, committed roads.

Play with the world

🍬 Food: click or drag to drop food (piles grow as you add). Sources deplete as they're carried away and shrink to nothing.
🪨 Rock: paint impassable terrain. Scroll on the canvas to resize the brush.
🏰 Nest: pick up the entire colony and drop it somewhere new; the old trails strand and the ants re‑route from scratch.
🧽 Erase: clear rock, scent, and food.
Scatter flings fresh food across the map; Evap / Wander / Speed retune the dynamics live; the trail‑layer toggles let you watch each pheromone on its own. Space pauses, R resets, keys 1–4 pick a tool.