In 2010, Atsushi Tero and colleagues placed oat flakes on a wet surface. The flakes marked the locations of cities surrounding Tokyo — Yokohama, Chiba, Sagamihara, the actual geography scaled to fit a petri dish. Then they released Physarum polycephalum at the point corresponding to Tokyo's center.
Over 26 hours, the slime mold spread outward. It found each food source. It extended tendrils, explored routes, connected nodes. And then — the part that became famous — it pruned itself back. Tubes carrying more traffic widened; underused connections narrowed and vanished. Twenty-six hours later, the network it had built was strikingly close to the actual Tokyo rail system: similar efficiency, similar fault tolerance, similar cost. The engineers who spent decades designing that railway had arrived at roughly the same solution as an organism with no brain.
The mechanism isn't mysterious, exactly, but it is strange. Physarum doesn't solve the problem by searching through possible routes. It doesn't evaluate options sequentially and pick the best. Instead, it occupies all options at once — physically extends itself into every available path — and then lets physics do the selection. Paths that carry more nutrients see their internal fluid oscillate more strongly, which widens the tube, which moves more fluid, which widens it further. Paths that are redundant or inefficient see less flow, contract, and disappear. The organism doesn't choose a solution; it becomes all the solutions and watches the right one survive.
This is different from search in a fundamental way. A classical computer is defined, in part, by its inability to be in two states at once — it evaluates alternatives sequentially. Physarum has no such constraint. It's a single cell with many nuclei, and it extends into all available paths simultaneously. The branching and feedback happen in parallel. The complexity of the problem doesn't accumulate linearly because the organism isn't iterating through possibilities; it's instantiating them all and running physics until convergence.
Researchers at Lanzhou University showed this explicitly: with focused light as negative feedback (Physarum is repelled by light, so illuminating a path makes it aversive), they set up a version of the Traveling Salesman Problem. With 8 cities, there are 2,520 possible routes. The organism found a feasible solution in time that grew only linearly as cities were added — a result that would be remarkable for any algorithm working through the same space sequentially. The oscillatory rhythm of the organism — regular contraction cycles that redistribute internal fluid — appeared to improve solution quality by introducing controlled variation into the search, like a biological equivalent of simulated annealing.
The most recent work adds a stranger layer. The organism appears to reach a non-equilibrium steady state where solution paths exhibit lower-frequency, larger-amplitude oscillations than non-solution paths — a measurable physical signature of correctness. There's also evidence of what's called Fröhlich condensation: a quantum phenomenon where molecular vibrations synchronize coherently across the organism, enabling long-range energy storage and communication. If that holds up under scrutiny, the slime mold may be doing something closer to quantum computation than classical search — using coherent oscillations to amplify the correct solution against the noise of wrong ones.
What's hard to locate in any of this is the boundary between computing and not-computing. The organism is doing chemistry. The chemistry has structure. The structure solves problems. At what point does structured chemistry become computation? The slime mold doesn't have that question. It has food, light, and the physics of fluid through tubes. We're the ones who look at the result and see the Tokyo subway — and call it brainless genius.