Life on the Cusp: The Trident G Theory of General Intelligence

Standard formulations of the free-energy principle treat adaptive agents as pure surprise minimisers, but in non-stationary niches this can drive brittle, energetically costly behaviour. This article proposes Trident G, a near-critical control architecture in which brains (and other intelligent agents) defend a non-zero metabolic–accuracy budget F★ and regulate a viable operating band on an asymmetric cusp. The state axis x = F − F★ tracks instantaneous deviation from this budget, while engagement ξ = G − F★, where G is expected free energy under the current generative model, and inference tilt δ (Convergent vs Divergent) function as control parameters that define three operational regimes: a subcritical Autopilot shaft, a near-critical Ψ-band that supports flow and fluid intelligence, and a supercritical Trident region with exploitation and exploration wells. General intelligence is reframed as the navigable volume 𝒢 of this landscape, decomposed into Control Range, Navigability (captured by a meta-efficacy η), and Niche Coupling (indexed by coherence χ and environmental feedback λ). I sketch how this cusp geometry can be instantiated from cortical columns to whole-brain control networks, and extended conceptually to teams and institutions, sketch falsifiable neural and behavioural signatures of Ψ-band occupancy and collapse, sketch falsifiable neural and behavioural signatures of Ψ-band occupancy and collapse, and argue that Trident G’s viable-band regulation offers a more resilient target for AI agents, cognitive training, and institutional design than pure free-energy minimisation. Keywords: free-energy principle, near-criticality, general intelligence, allostasis, cusp catastrophe, cognitive resilience