Activator–inhibitor systems compute best near regime transitions, not in stable patterns.

Scientists often study pattern-forming systems, like chemical mixtures that make stripes or spots, just to see the shapes they create. However, the same rules that make patterns can also process information. In this study, researchers built a grid-based computer model inspired by chemical reactions. The model uses two kinds of rules: a smooth curve and a step-like rule. By turning one dial in each rule, they changed the system's response without changing how grid cells connect. This shift pushed the grid through four states: collapsed, structured, saturated, and overshooting. Next, the team tested the system's memory and ability to tell inputs apart. They found that the model computes best right near the transitions between these states, rather than inside the states themselves. This reveals that materials that make patterns can also process data by slightly tweaking their local behavior.

Why it matters

For engineers, this suggests new ways to build computers from physical media instead of silicon. For biologists, it hints that pattern-forming tissue might already be doing computation. The key insight is that the computational sweet spot is at the transition between regimes, not in stable states.

Questions answered