Teleodynamic closure and the definition of intelligence

It has been proposed that intelligence is "an emergent property of consciousness … the ability to intentionally solve challenges and adapt to new situations and to an ever-changing environment," anchored in the far-from-equilibrium (FFE) thermodynamics of intelligent systems (Vitas, Cvjetović and Dobovišek, BioSystems 263, 105776, 2026; hereafter VCD). We argue that this proposal, while moving in a productive direction, omits the specific dynamical regime that distinguishes intelligent organisation from mere dissipative complexity: the teleodynamic regime, in which a system's constraints participate in their own generation through recursive, constraint-on-constraint dynamics - teleodynamic closure - and thereby produce end-directedness from non-intentional substrates. Without that layer, the central operative terms of the definition - "intentionally," "adapt," "challenge" - remain undischarged. We identify five lacunae: (i) the FFE invocation is thermodynamic but not yet teleodynamic; (ii) intentionality is presupposed rather than emergentist-derived; (iii) the proposed cognition→consciousness→intelligence ordering is misaligned with the aneural-cognition literature; (iv) "adaptation" without a constraint-generation account is operationally indistinguishable from passive equilibration; (v) the definition lacks any treatment of constraint plasticity. We propose a reconstructed definition: intelligence is the capacity of a teleodynamically organised system to generate, propagate and modify the informational-regulatory constraints of its cognitive subsystem such that the system's own end-directedness is preserved and extended across novel adaptive challenges in a far-from-equilibrium environment. Building on the regulatory tradition (Bich et al., 2016; Bich and Moreno, BioSystems 148, 12-21, 2016), we distinguish life (closure of constitutive constraints), cognition (the informational-regulatory subsystem controlling those constraints), and intelligence (the constraint-plastic mode of that subsystem). The reconstruction is grounded in a timescale-separated dynamical sketch and worked biological cases (chemotaxis and metabolic switching in E. coli). It is substrate-independent yet thermodynamically anchored, consistent with aneural cognition, independent of any prior commitment to consciousness, and operationalisable for artificial intelligence, astrobiology and biological cognition research.