Knowledge systems — genetic, cultural, computational — are dissipative structures in the sense of Prigogine: they hold local order only while exporting entropy to their environment. A cell, a library, a reasoning machine — each persists because it is burning free energy against the drift toward disorder. The bill is physical.
For an inference substrate, that is the constraint that binds. Gentian's SRAM-first substrate is the architectural answer: don't shuttle state across a memory boundary, and the Joules follow.
Notes
Knowledge and entropy — from living systems to AI
Thermodynamic framing of knowledge amplification: Gibbs free energy, Shannon entropy reduction, dissipative structures, Landauer's principle.
Historical evolution of knowledge systems
Epochs of knowledge evolution — genetic inheritance, social learning, language, writing, print, digital systems, AI — analysed through the same thermodynamic lens.