No ant has a map
An ant colony finds the shortest path to food. No ant knows the shortest path. No ant has a map. No ant is in charge. Each ant follows two rules: (1) follow pheromone trailsChemical signals deposited by ants that evaporate over time. Shorter paths get reinforced faster (more ants traverse them before the pheromone fades). This is how the colony ‘computes’ shortest paths., (2) leave pheromone behind.
From those two rules, the colony computes optimal foraging routes. The computation is distributed across thousands of simple agents. No individual agent computes anything useful. The system computes.
This is Skeleton 4: distributed computationInformation processing spread across many simple agents with no central controller. The computation is a property of the system, not any individual component.. Simple local rules producing global information processing without any central controller.
Four rules, infinite complexity
Conway's Game of Life. A grid. Each cell is alive or dead. Four rules:
Rules
- Alive + fewer than 2 neighbors → dies (underpopulation)
- Alive + 2 or 3 neighbors → lives
- Alive + more than 3 neighbors → dies (overcrowding)
- Dead + exactly 3 neighbors → becomes alive
What emerges
GlidersA pattern in Conway’s Game of Life that translates across the grid over successive generations. The simplest moving structure, and a building block for constructing logic gates. (move across the grid). Oscillators (repeat). Glider guns (produce gliders). Logic gates. A Turing-completeCapable of simulating any computation that a Turing machine can perform, given enough time and memory. The theoretical ceiling of computational power. computer. From four rules about counting neighbors.
Left: a glider (moves diagonally). Right: a blinker (oscillates in place).
The key insight: the rules are trivial. The behavior is not. The gap between rule complexity and behavior complexity is where computation lives.
Edge of chaos = the critical point
Not all rule sets produce interesting computation. Most produce one of two outcomes:
Ordered
Everything freezes. Information can't flow. Structures are static. No computation.
Critical
Long-range correlations. Information flows freely. Complex structures form AND dissolve. Computation concentrates here.
Chaotic
Everything fluctuates. Information is destroyed by noise. No stable structures. No computation.
Langton (1990)1 showed this explicitly: computational capability peaks at the phase transitionA sudden qualitative change in behavior at a threshold value. Water freezing, magnets demagnetizing, neural networks shifting from ordered to chaotic. The critical point is the boundary. between order and chaos. Too ordered = frozen, can’t compute. Too chaotic = noisy, can’t remember. The sweet spot is the critical point.
The bridge to Skeleton 2
This is where Skeleton 4 (Computation) meets Skeleton 2 (Dynamics/Criticality) — the B–D bridgeBridge letters: A = Topology, B = Criticality, C = Autocatalysis / Closure, D = Computation. B–D links Criticality and Computation. (Criticality ↔ Computation). The edge of chaosThe boundary between ordered (frozen, predictable) and chaotic (noisy, unpredictable) behavior. Where computation is maximized — ordered enough to store information, chaotic enough to process it. coincides with the critical point: computation concentrates at criticality. The correspondence is close and quantitative — the same mathematics viewed from two directions — though whether the two are literally identical or a tight structural analogy is exactly the kind of claim this program is still testing.
Computation adds a new lens to criticality: the critical point isn't just where phase transitions happen. It's where information processing is maximized. The system computes most powerfully at exactly the point where it's about to change state.
Two depths of computation
Depth 1: Universality
The system can simulate any computation, given enough time and space. Conway's Life is Turing-complete. Random Boolean networksNetworks where each node has a binary state (on/off) and updates based on logical rules applied to its inputs. Kauffman used them to model gene regulatory networks. at criticality are Turing-complete. A system at depth 1 has the capacity for arbitrary computation.
Universality is about potential: the computational substrate is powerful enough, in principle, to compute anything. Whether it actually does is a separate question.
Depth 2: Coordination
The system doesn't just have computational capacity — it uses it to coordinate behavior across scales. 86 billion neurons don't just fire randomly at criticality. They coordinate into perception, decision, action. Millions of ants don't just diffuse randomly. They coordinate into foraging, nest building, defense.
Coordination is universality harnessed. The gap between "can compute" and "does compute something useful" is the gap between depth 1 and depth 2.
Many systems have universality (depth 1). Fewer achieve coordination (depth 2). A cellular automaton at criticality has universality. A brain at criticality has coordination. An ML network at double descent has coordination. The difference is: depth 2 systems have functional computation, not just potential computation.
The brain computes at criticality
Neuronal avalanches follow power lawsMathematical patterns with no characteristic scale. Small events are common, large events are rare, and the ratio between scales is constant. The signature of criticality.. This was one of the early criticality observations (Beggs & Plenz, 2003)2. The brain operates near a critical point.
But the brain doesn't just exist at criticality. It computes there. The critical point maximizes:
Dynamic range
Dynamic rangeThe range of input intensities a system can distinguish. At criticality, dynamic range is maximized — the system can detect both whispers and shouts without saturating.: the range of stimulus intensities the system can distinguish. Maximized at criticality. Subcritical systems compress dynamic range. Supercritical systems saturate.
Information transfer
Mutual informationA measure of how much knowing one variable tells you about another. Between distant neurons, mutual information peaks at criticality — the critical point enables long-range communication. between distant neurons. Maximized at criticality. The critical point is where correlations span the whole system, enabling long-range coordination.
The brain has both depth 1 (universality — it can, in principle, compute anything) and depth 2 (coordination — it actually coordinates perception, thought, and action). It's the prototypical Skeleton 4 system.
Eight fields, one framework
Distributed computation was discovered and rediscovered across at least eight fields:
Universal computation. Abstract machines. Proved limits of computability.
Cellular automata. Self-reproducing machines. Showed computation could be distributed.
Game of Life. Proved simple local rules can produce universal computation.
Self-organized criticalityA system that naturally evolves toward a critical state without external tuning. Bak, Tang, and Wiesenfeld (1987) showed sandpiles do this — the brain may too.. Systems that tune themselves to the computational sweet spot.
Edge of chaos. Explicitly showed computation peaks at the phase transition.
Boolean networks. Random gene regulatory networks compute at criticality.
Echo state networks. Showed that any sufficiently complex dynamical system near criticality can compute.
A New Kind of Science. Classified all simple programs. Class 4 (between order and chaos) = universal computation.
Mathematics → automata → game theory → statistical physics → complex systems → biology → machine learning → computational theory. Eight approaches to the same insight: complex computation arises at the boundary between order and chaos.
Recap
- Skeleton
- Distributed computation / universality — simple local rules producing global information processing without central control.
- Depth 1: Universality
- The system has the capacity for arbitrary computation. Turing-complete. The substrate is powerful enough.
- Depth 2: Coordination
- The system uses its computational capacity to coordinate behavior across scales. Perception, foraging, learning. Computation harnessed.
- Bridge to Skeleton 2
- Edge of chaos = critical point. Computation concentrates at criticality. This is the B–D bridgeBridge letters: A = Topology, B = Criticality, C = Autocatalysis / Closure, D = Computation. B–D links Criticality and Computation. (Criticality ↔ Computation): Skeleton 2 and Skeleton 4 as closely related mathematics seen from two directions.
- Key insight
- You can't compute in a frozen system (no information flow) or a chaotic one (no memory). The critical point is where computation lives.
- Convergence
- 8 fields independently arrived at distributed computation at the critical point. Turing, von Neumann, Conway, Bak, Langton, Kauffman, reservoir computing, Wolfram.