Knowledge is typically treated as an abstract cognitive achievement-something minds possess, acquire, and trade. This paper argues for a different view: knowledge is a specific thermodynamic state, one that certain sufficiently complex dissipative structures reach when their substrate conditions are met. Drawing on Prigogine's dissipative structures, Schrödinger's negentropy, and Friston's free energy principle, I propose that the transition from information to knowledge is a phase change-a qualitative reorganisation from lookup table to generative model. The biological substrate of this transition is the dual-circuit architecture of the mammalian brain: fast ionotropic receptors handle moment-to-moment traffic while slow metabotropic receptors remodel the network's underlying topology, compressing accumulated experience into generalisable principles. The conscious experience of insight is the registration of a compression event that has been accumulating subthreshold. If knowledge is compressed predictive modelling, then the compression ratio (environmental complexity ÷ model size) is a principled proxy for knowledge depth, dissolving binary debates about whether a system "really" knows into questions of degree. The boundary between biological knowledge and what artificial systems do with information may be a matter of compression ratio rather than category membership.
Robert Keith Russell (Sat,) studied this question.