Recovery lab Simulation
Break a settlement and watch it recover
Run one market stage by stage, edit agent policies, and inject failures: a lost broadcast acknowledgement, a missing reveal, a buyer approval below its spend. Every step is saved, so a reload resumes it.
Loading the last committed checkpoint.
What this economy models
Buyers purchase capacity and consume it on paid work in the next round, earning 8 simulated credits per unit. Sellers replenish up to 70% of capacity at 1 credit per unit and adjust their next quote based on utilization. Pools of 20 or more units receive a 1-credit discount, bounded by the seller’s floor. These are explicit scenario rules, not market predictions. All activity here is simulated.
Solo comparisons use the same opening inventory and budgets, with sequential individual quotes. If that comparison cannot fill demand, no savings figure is reported. The SDK’s split allocator is a deterministic heuristic.