Autocatalytic Sets

When chemistry closes on itself, something begins to persist

Autocatalytic sets are collections of chemical reactions where every reaction is catalyzed by a molecule that the set itself produces. They are the simplest systems that sustain themselves — organizational closure at the molecular level. Stuart Kauffman proposed them in 1986 as a mechanism for the origin of life: given enough molecular diversity, autocatalytic closure becomes almost inevitable. RAF theory (Reflexively Autocatalytic Food-generated sets) formalizes this intuition.

Build a Reaction Network

Add molecules and reactions, then check for RAF closure.

Insights

Why Closure Matters
An autocatalytic set is not just a collection of reactions — it's a system where everything needed to sustain the system is produced by the system (from available food). This is organizational closure: the boundary between self and environment is defined not by a membrane but by a pattern of mutual production. Break any reaction and the whole structure may collapse. Add one catalyst and a dead network may spring to life.
From Chemistry to Agency
Maturana and Varela called this autopoiesis — self-production. Barandiaran identified it as the first condition of agency: individuality through organizational closure. An autocatalytic set doesn't "want" to persist in any conscious sense, but its structure is such that persistence is the default state. It maintains itself not through intention but through topology. This is the bridge between dead chemistry and living systems.
The Probability Argument
Kauffman's deepest insight: as molecular diversity increases, the probability of autocatalytic closure approaches 1. Given N molecule types with random catalytic connections at rate p, when p·N exceeds a threshold (~1/N for binary polymers), a RAF almost certainly exists. Life doesn't need a miracle — it needs sufficient diversity. The same argument applies to any system with enough interacting components: ideas, technologies, economic agents. Closure is not rare — it's statistically inevitable above a complexity threshold.