guruswami-ai

Flockinator is a hands-on laboratory for exploring how small, resource-constrained devices can learn and cooperate as a swarm. Sixteen ESP32 nodes form a physical display of a shared simulated world, with an ESP32 controller hosting the simulation, web controls and live dashboard. The goal is to understand how individual decisions and communication contribute to collective intelligence and what happens when information is delayed, missing or misleading. The project addresses the difficulty of making multi-agent learning observable and measurable. Its experimental workflow pairs each node with a Meta Muse reviewer, coordinated by Muse Wonder Full, to assess recorded decisions and communications and propose improvements. Over repeated trials, we aim to compare shared and specialised firmware, measure whether changes improve performance, and test whether learning under adverse conditions produces more capable, resilient behaviour.