A robot in a maze steps into a dead end, and a minute later finds the exit. No mistake: that’s the cartographer’s tactic. It circles the mountain from all sides to draw an accurate map. Scout algorithms now do the same: instead of chasing a single goal, they collect a set of different routes.

The new approach, mimicking evolution, launches several seekers at once with small variations. Their trajectories repel each other like magnets — to avoid duplication. Discoveries go into a shared archive, where each solution stores its code and visible behaviors (such as coordinates). In maze tests, such teams gathered up to 3199 unique maneuvers. Versus 2824 for soloists.

The method will be useful for computer models and stochastic tasks, where you need not one answer but a spectrum of options: from drug design to robot plans. The uncharted continent is now mapped not by lone explorers but by coordinated teams. Every sidestep is a road to discovery.