Haueis and Colaço’s target article is timely and important. At a moment when artificial intelligence, neuroscience, and the intersection of the two (i.e., NeuroAI) increasingly position themselves as the primary explanatory domains for cognition, the authors’ insistence that metabolic considerations constrain and generate cognitive models represents a necessary corrective. Their central claim – that cognition cannot be adequately modeled without accounting for the brain’s energetic limitations and organization – directly challenges the long-standing emphasis on abstraction in cognitivist and other computational-oriented approaches that treat information processing as effectively costless. Though there is much to laud about this approach, there remains a glaring omission: time, specifically, multiscale dynamics. While the authors discuss dynamics a bit, temporality does not appear to be central to modeling or explaining cognition for them. The main aim of this commentary is to motivate the idea that temporality – specifically in the form of multiscale dynamics – ought to be treated as fundamental to understanding cognition, especially in biological systems. Accordingly, Haueis and Colaço’s already-compelling project can be made more so by incorporating such temporal features.

A long-standing criticism of computational accounts of cognition is their atemporal nature (Brette, Reference Brette2026; Favela & Raja, Reference Favela and Raja2026; van Gelder, Reference Van Gelder1995). Here, “computational” refers to the idea that cognition is a form of information processing along the lines of a Turing machine (i.e., the algorithmic manipulation of symbols; Rescorla, Reference Rescorla, Zalta and Nodelman2025) and that brains instantiate such processing in some substantive way (e.g., Churchland & Sejnowski, Reference Churchland and Sejnowski1992/Reference Churchland and Sejnowski2017; Rolls, Reference Rolls2021). This limitation has prompted attempts to reorient various relevant disciplines in ways that place temporal features at more central explanatory positions, including cognitive science (e.g., Ward, Reference Ward2002) and neuroscience (e.g., Izhikevich, Reference Izhikevich2007). This shift has often been accompanied by increased attention on the brain’s multiscale spatial organization (e.g., D’Angelo & Jirsa, Reference D’Angelo and Jirsa2022; Haueis, Reference Haueis2022; Thomas, Forrester, & Ronald, Reference Thomas, Forrester and Ronald2016), as well as its multiscale dynamics (e.g., Munn et al., Reference Munn, Müller, Favre-Bulle, Scott, Lizier, Breakspear and Shine2024; Siettos & Starke, Reference Siettos and Starke2016; Zhang et al., Reference Zhang, Kalies, Kelso and Tognoli2020). One consequence of such disciplinary expansion upon the more traditional approaches is that cognition is becoming viewed as fundamentally a temporally extended, multiscale dynamical process. From this perspective, cognition is understood as emerging from the coordination of processes across multiple spatial and temporal scales, from single neurons to organism–environment interactions (Favela, Reference Favela2024). In such a framework, metabolism is not an isolated constraint but is itself embedded within – and subordinate to – ongoing dynamical organization. Energy consumption, allocation, and efficiency are always in the service of maintaining and adapting these coordinated dynamics in order to cause and constitute various cognitive processes.

This perspective reveals a deeper issue: metabolism without time risks reifying static constraints rather than capturing living processes. Haueis and Colaço treat metabolic factors as bounding and shaping computation, but they do not fully integrate the idea that these constraints are themselves dynamically modulated across timescales. For example, gene expression, synaptic plasticity, neural oscillations, and large-scale functional connectivity all unfold over distinct temporal regimes, yet must be coordinated for cognition to occur. Even metabolic processes – such as ATP production or glucose utilization – are temporally structured and context-sensitive, fluctuating with task demands and environmental coupling. Thus, from a multiscale dynamical viewpoint, metabolism should not be treated as a mere constraint on models; it should be understood as participating in temporally nested systems of interactions. The brain’s energy use is not simply a budget to be respected, but a dynamically allocated resource that co-evolves with cognitive activity – not to mention nonneural physiology, whole bodies, and features of the environment. In terms of doing cognitive science with an emphasis on brains, explanations shift to tracking how patterns evolve over time (e.g., Buzsáki, Reference Buzsáki2006; Criscuolo, Schwartze, & Kotz, Reference Criscuolo, Schwartze and Kotz2022) rather than identifying static components (e.g., representations) or computational constraints (e.g., algorithms).

Importantly, incorporating multiscale dynamics would strengthen, rather than undermine, Haueis and Colaço’s project. Their evaluative and generative framework could be extended to include temporal constraints, where models would need to not only fit within energy budgets but also to reproduce the timing, coordination, and stability properties of real (i.e., biological) cognitive systems. For instance, a metabolically plausible model that fails to capture the temporal synchronization of neural populations during perception-action cycles would remain explanatorily deficient. Furthermore, emphasizing dynamics helps bridge the gap between metabolism (i.e., molecular) and more situated, organism-environment scales. While Haueis and Colaço focus on neural and computational modeling, other work that aims at more system-level targets of investigation aim to situate cognition within continuous organism–environment interactions, where timing and coordination are essential (e.g., Criscuolo, Schwartze, & Kotz, Reference Criscuolo, Schwartze and Kotz2022; Kelso, Reference Kelso2012; additional examples in Favela, Reference Favela2024). Metabolic processes, in this context, support the organism’s ability to maintain these interactions over time. Thus, metabolism is best understood as enabling the ongoing dynamical coupling that constitutes cognition, rather than merely constraining internal computations.

In conclusion, Haueis and Colaço’s article is a timely and important contribution that rightly foregrounds metabolic considerations in cognitive modeling, particularly in an era dominated by computational-centric approaches (cf. Favela, Reference Favela, Kelty-Stephen and Mangalam2026; Favela & Amon, Reference Favela and Amon2023). Their work pushes the field toward greater biological realism and offers valuable tools for constraining and generating models. However, from the perspective of more systems-level approaches (e.g., Chemero, Reference Chemero2009; Favela, Reference Favela2024; Pessoa, Reference Pessoa2022), their framework remains incomplete without a more explicit and central role for multiscale temporal dynamics. Biological cognition is not just energy-limited; it is dynamically organized across time, and metabolism itself is embedded within this organization. Recognizing this would not only address a key omission but also further enhance the explanatory power of their approach.