minAction.net

Living and learning systems organise toward a single functional.

The Network-Weighted Action Principle (NWAP) proposes that biological networks, neural architectures, and physical-law-discovery systems all extremise the same quantity — a balance of energy minimisation, information maximisation, and connection-cost penalty:

The empirical prediction is sharp and falsifiable: networks under these constraints organise toward high modularity, with the modularity excess over appropriate null models scaling with optimisation pressure, and the training-energy required to learn falling under the same objective in engineered systems.

Over 2023–2026 we tested this prediction in four independent domains — physiology, physics, neural architecture, and biology — and the modularity-excess and energy-efficiency signatures appear in each. This site is the integrated record of that programme.

The causal problem

The classical view in biology and machine learning is horizontal: each scale has its own mechanism, identified by within-scale interventions. Modern physiology has worked this way for a century; machine learning has worked this way since prediction error became the dominant optimisation target. But neither tradition has produced a vertically organising principle — a law that connects scales, that predicts what kind of structure should emerge under shared constraints.

NWAP is proposed as such a principle. It does not replace within-scale mechanisms; it predicts the architectural target that within-scale mechanisms converge on under the dual constraint of energy and information.

Life at the edge of chaos

Living systems sit in a narrow phase-transition zone between rigid order (crystals, ischaemia) and dissipative chaos (asthma, hyperinflammation). NWAP picks out this zone as the equilibrium of the action functional — neither pure energy minimisation nor pure information maximisation, but the joint extremum that admits both stability and adaptability.

Key evidence — modularity emerges from cost minimisation

Connection-cost minimisation, formally derived by Clune et al. (2013) and shown in Frasch 2026a, Figure 1D, produces modularity as the structural signature: networks partition into densely-connected communities with sparse inter-module connections. The 2026 papers test this prediction in their respective domains.

"Meaning" — the speculative payoff

At the intersection of the four neighbouring variational frameworks (free-energy, dissipative adaptation, constructal theory, and the Network-Weighted Action) sits a thought-provoking conceptual target: meaning, operationally defined as successful uncertainty reduction through efficient action. Each framework captures one aspect of this convergence; NWAP is unique in carrying it as a measurable architectural prediction. We treat this as an open question worth posing, even if the current data cannot yet discriminate among the four accounts.

"Meaning" — operationally defined as the successful reduction of uncertainty through predictive modelling and efficient action — has the same shape at every scale of biological organisation:

The four-domain validation, at a glance

Read the framework paper →  ·  Browse the four validations →  ·  Future work →