CONTROLAgents & automation
Entropy City
Farynth@farynth
2:43Entropy City is a simulation where an AI named Jev coordinates disaster response and recovery decisions for a city experiencing cascading failures.
SignalIt chooses who should respond, where they should go, and which work should take priority.
I’ve been experimenting with @typesafeai 's jev, and one idea really caught my attention: could a fast AI coordinate the response to a city gradually falling apart?
So I built Entropy City.
Entropy City starts with no AI management.
Supplies run down. Infrastructure wears out. Fires spread, storms roll through, and broken systems start putting pressure on everything connected to them.
Then I give Jev control of dispatch and recovery decisions.
Jev receives a snapshot of the city: damaged infrastructure, supply levels, company balances, available crews, and their current assignments. It chooses who should respond, where they should go, and which work should take priority.
Repair the power supply. Send crews to a fire. Get water and food moving. Clear snow-covered roads. Preserve resources by pausing less urgent work.
The simulation executes those decisions, updates the world, and sends the next snapshot back to Jev. That creates a continuous feedback loop: observe, decide, act, repeat.
Getting this working exposed some interesting problems.
Request timing, traffic bottlenecks, and how we presented the available choices all mattered.
Fast inference only helps if the surrounding system can turn decisions into useful action.
It’s an early prototype, but I can see where this could lead.
Utilities, logistics networks, factories, maintenance fleets, systems with thousands of connected operations that constantly need to adjust as conditions change.
Honestly the more I think about the possibilities this unlocks, the more I am amazed.