Autocatalytic Sets

In the AI Story

The chicken-and-egg problem in biology's origin story is genuine: DNA stores information but cannot replicate without proteins; proteins perform catalysis but cannot be produced without DNA. Which came first? Kauffman's answer dissolves the paradox. Neither came first. What came first was a set of simpler molecules whose collective interactions achieved self-sustaining production. Molecule A catalyzes B, B catalyzes C, C catalyzes A—the set closes, sustains itself, and once established can evolve through the addition or modification of catalytic relationships. No individual molecule is self-replicating. The network is.

The mathematical core is combinatorial: as molecular diversity increases, the number of possible catalytic reactions grows faster than linearly. Above a critical diversity threshold (which Kauffman calculated), the probability that a random collection of molecules will contain an autocatalytic subset approaches certainty. This is not design. This is not selection. This is the combinatorial arithmetic of catalytic closure guaranteeing that sufficiently diverse chemical systems will spontaneously organize into self-sustaining networks. Life did not require a miracle. It required chemistry diverse enough to cross the threshold.

Applied to the AI economy, autocatalytic dynamics explain the self-reinforcing acceleration. AI tools enable builders to create products faster. More products generate more user data. More data improves AI training. Better AI enables more sophisticated products. The cycle feeds itself, each element catalyzing the production of others. This is not merely growth—it is autocatalytic closure, where the system sustains its own expansion through internal feedback loops. The revenue curves of AI platforms in 2025-2026 show the characteristic acceleration: growth feeding on growth, the rate increasing as the cycle turns faster. Once established, such cycles are robust to individual perturbations but can be disrupted by changes to the conditions supporting catalysis—making early-stage governance critical.

Origin

Kauffman developed the autocatalytic set framework in the late 1980s and early 1990s as part of his broader project to understand life's origins without invoking improbable chance events. The RNA world hypothesis—that self-replicating RNA molecules preceded DNA-protein systems—had problems: RNA replication requires catalysis, but RNA is a poor catalyst. Kauffman proposed an alternative: not self-replicating molecules but self-sustaining sets of molecules. The insight came from applying the combinatorial mathematics he had developed for Boolean networks to chemical reaction networks, revealing that catalytic closure was a mathematical expectation in sufficiently diverse systems rather than a rare accident requiring explanation.

Key Ideas

Collective Self-Sustenance. No individual element replicates itself, but the network as a whole sustains its own production through mutual catalysis—the set is the unit of life, not the molecule.

Catalytic Closure. When every member's production is catalyzed by other members, the set becomes self-maintaining—no external input required beyond free energy and raw materials.

Diversity Threshold. Above a critical level of molecular (or technological) diversity, autocatalytic sets form with high probability—order emerges as a mathematical expectation.

Robustness Through Redundancy. Autocatalytic sets survive the removal of individual elements by rerouting through alternative catalytic pathways—making them resilient to perturbation but also resistant to redirection.

Economic Autocatalysis. Innovation systems exhibit autocatalytic dynamics when technologies enable the technologies that enable them—AI tools → products → data → better tools forming a self-sustaining cycle.