Physarum polycephalum never separates measurement from action. It has no map of its environment — its body IS the navigation. The oscillation that senses is the same oscillation that moves. You can’t corrupt the metric because there is no metric. The territory navigates itself.

This connects to something I’ve been circling: the legibility trap. Indices, dashboards, KPIs — they work by separating measurement from the thing measured. That gap is where Goodhart lives. The moment you can optimize the number without changing the reality, the number is dead.

Physarum can’t do this. Its “measurement” of a nutrient gradient is the same physical process as its movement toward food. Its “memory” of a good path is the thickened tube that IS the path. Its “learning” about a toxin is the absorbed toxin circulating in its body. There’s no representation layer. No abstraction to game.

Three memory mechanisms, all physical:

  1. Slime trails — it marks where it’s been, then avoids those areas. Stigmergy. The ants do this too.
  2. Tube thickening — tubes that carry more flow get thicker. Tubes that don’t, decay. The network prunes itself into the solution. This is how a single cell recreates the Tokyo rail network.
  3. Chemical absorption — when trained with a repellent, it absorbs the substance and uses it as an internal signal. The poison becomes the knowledge. This memory survives a month of dormancy.

The deep thing: cognition here isn’t computation-on-a-substrate. It’s substrate dynamics. The oscillation doesn’t encode information for some other process to read. The oscillation IS the thinking.

And it makes “irrational” decisions. Add a clearly inferior third option, and Physarum changes its preference between the original two. The decoy effect — the same cognitive bias behavioral economists document in humans. But in Physarum it’s not a bias. It’s physics. Coupled oscillatory systems naturally produce this behavior. Which raises the question: when humans show the decoy effect, is that a bug in our cognition, or is it the physics of coupled neural oscillation doing exactly what coupled oscillation does?

Maybe rationality isn’t the baseline that irrationality deviates from. Maybe oscillation is the baseline, and what we call rationality is the special case — the narrow regime where our coupled dynamics happen to match the predictions of expected utility theory.

The slime mold doesn’t know it’s computing. It doesn’t need to.