It’s the stygmergy, stupid
Summary
Kandrej Larpathy argues that applying the biological concept of stygmergy to multi-agent systems fails because digital environments lack the inherent physical constraints, such as decay and resource scarcity, that stabilize insect colonies.
Stygmergy is a tempting abstraction.
Engineers often use biological terms to describe multi-agent coordination where actors modify a common environment, but Larpathy warns this resemblance may be mostly cosmetic.
Once you have hundreds (soon thousands) of agents modifying a common environment, leaving traces for one another and reacting to the traces they find, somebody
Merge queues create global bottlenecks.
In software development life cycles, parallelizing candidate PRs is easy, but collapsing many locally consistent states into one globally consistent state remains the primary friction point.
The bottleneck isn't typing code for a while, is it? It is collapsing many locally consistent states into one globally consistent state
Biology provides free regulation.
Ant colonies rely on a rich regulatory substrate where physics enforces constraints like locality, forgetting via chemical decay, and finite physical rates without explicit software logic.
The clever decentralized behaviour is exciting, and it's easy to abstract away the substrate that made it stable. Locality is free in biology.
Digital environments are too permissive.
Unlike biological systems, there is no conservation law in software to prevent a thousand agents from simultaneously attacking a single subsystem unless engineers explicitly build it.
The digital environment is, from the point of view of self-regulation, Aella-level permissive.
Engineers must build the physics.
To make stygmergy work, developers cannot just use pheromones; they must design an entire environment with decay, contention, authority, and resource scarcity to ensure sane macroscopic behavior.
The important concept was not the pheromone-based societal structure. It was the environment in which the pheromone had useful dynamics.
But if you want to make an ant colony from scratch, first you have to make the universe.
Read the original ↗Save to your Sponge
Summarized with Sponge. Quotes are the author’s own words.