August 15, 2026

Provisional research programme roadmap

1. Wattage of learning: LTP rate per CMRO₂
   1. Synapses to policies: from energy constraints on synaptic change to energy constraints on behavioral change
2. Rate-limited policy update: policy change rate per CMRO₂, including maintenance and suppression of alternative policies, where a “policy” is parameterized by synaptic weights, gains, precisions, etc.
   • Energy-weighted expected utility
   • Energy-return on energy-invested (EROEI)
   • Multi-reward Hebbian landscape perturbation
   • Metabolic neuroeconomics
   1. Disease models:
      • Addiction: short-term free-energy minimization at the cost of long-term free-energy minimization? Low confidence, short discounting? High confidence, long discounting?
      • Autism: preference for very expensive, very accurate internal generative models?
      • Schizophrenia: opposite? Metabolically frugal?
      • ADHD: low cost of policy change?
      • Depression: high cost of policy change?
   2. “Neurometabolic rift”: describes persistent incongruities between energy returns vs. investment despite organismal adaptation, cf. John Bellamy Foster’s “metabolic rift,” schizoanalysis’ “anti-production,” and theories of [mental] health and disease.
   3. A thermodynamically incentive-compatible decision theory, viewing animals as dissipative structures selected to behaviorally adapt for EROEI and entropy-export
      • Rate-limited policy update puts the empirical number on the wattage
3. Reverse game-theoretic mechanism-design to reconcile the neurometabolic rift