An onboard AI is dropped into an unexplored star system full of asteroids and moons, each with its own resource value, scientific value and hazard risk, and has to plan a single probe's exploration route under a hard fuel budget. Rather than brute-forcing every possible visiting order — the orienteering problem this models is NP-hard — it uses the same greedy utility-density heuristic real mission-planning tools rely on: at every step, chase whichever reachable target offers the best weighted payoff per unit of fuel spent getting there, and stop the moment the budget runs out. Tune how much the AI values resources versus science versus its own safety and watch the whole strategy — and the route it draws through the system — change in real time.