OpenSky · vehicle-compute network

A fleet of vehicle-mounted Raspberry Pis harvest solar + wind while driving, run LLM inference + PoET proofs, and earn coins. The binding constraint is the per-unit energy budget — battery = health, running = drain, harvesting = capacity-limited replenishment, shutdown = stranding. The hierholzer_v scheduler plans the fleet; an economics layer turns scheduled uptime into coins → revenue → P&L, swept over device count N.

Levers

Device fleet ·

Each row is a device over 24 h. computing · solar harvest · wind harvest · stranded (hit shutdown voltage). The line is battery state-of-charge.

Scale-out projection · the product number

The engine schedules a representative fleet; the economics extrapolate across device count N (harvest capacity scales with N, so per-device energy behavior is N-invariant). Watch the difficulty-damping bend past ~100 active devices, and whether the fleet pays back within its 36-month life.