Problem Space Search: Hill Climbing vs Simulated Annealing

Newell and Simon modelled problem solving as search through a "problem space" — states connected by operators, moving from a start toward a goal. This simulator renders that idea literally: a real 3D landscape built from overlapping Gaussian peaks, with a searching agent that can run greedy hill climbing, random-restart hill climbing, or simulated annealing across it. Watch hill climbing get trapped the instant it reaches a local maximum, watch random restarts try again from a fresh state, and watch simulated annealing's cooling temperature let it tolerate temporarily worse moves — a direct computational parallel to the incubation and functional-fixedness effects the accompanying article covers — before settling onto the true global optimum.