CALLIOPE · BCOM

A 25-year research program that asks: if brains are compression engines, what does that tell us about consciousness, emotion, and mental illness?

The central bet is simple but radical. Everything a brain — or any cognitive system — can ever know arrives as a stream of data. The only thing it can do with that data is find patterns and compress them into a model. Kolmogorov Theory (KT) takes this seriously as a foundation, not just a metaphor. It borrows Kolmogorov complexity (the length of the shortest program that reproduces a dataset) as its core measuring stick, and then builds upward: cognition is model-building, structured experience arises when a model is successfully matched against incoming data at an internal "Comparator," and the richness of conscious experience tracks how good — how compressive and encompassing — that model is.

The agent architecture is the load-bearing structure. KT defines an algorithmic agent as having three modules: a Modeling Engine (builds and runs compressive world-models), a Planning Engine (runs counterfactual simulations to choose actions), and an Objective Function (evaluates states against homeostatic goals and produces valence — the felt sense of good or bad). Emotion, in this framework, is not a separate faculty but the full tuple of (model, valence, plan). The classical dimensions of emotion — cognitive content, hedonic tone, arousal — map one-to-one onto the three modules. This is a genuinely structural claim, not a loose analogy.

The mathematical spine tightens considerably in the later papers. A 2026 result (the "Algorithmic Regulator") proves, without assuming any probability distribution, that a system which effectively regulates its environment must share algorithmic structure with it — a distribution-free, single-sequence version of the old cybernetic principle that "every good regulator must contain a model of the system it regulates." A companion paper grounds this geometrically: when the world's data is generated by a Lie group of symmetry transformations, an agent that tracks the world must evolve on reduced invariant manifolds that mirror those symmetries. Conservation laws and dimensionality reduction fall out as consequences, not assumptions.

The empirical work is not decorative. LSD fMRI data analyzed through Ising models shows that the drug raises the brain's effective "temperature" (a measure of disorder in the statistical-physics sense) and increases algorithmic complexity — consistent with KT's prediction that disrupted modeling inflates apparent complexity. A laminar neural mass model (LaNMM) is used to simulate Alzheimer's disease progression: progressive loss of fast inhibitory interneurons reproduces the biphasic EEG trajectory of AD (early hyperexcitability, then oscillatory slowing), and simulated serotonergic psychedelic action partially reverses it. A separate modeling study proposes that cross-frequency coupling — slower brain rhythms modulating the amplitude of faster ones — is the circuit-level mechanism for hierarchical prediction-error evaluation, with a predicted logarithmic spacing of oscillation bands that matches the observed delta-through-gamma structure.

KT is not trying to solve the hard problem of consciousness — why there is experience at all. It takes experience as axiomatic and asks how it acquires structure. That is a deliberate and honest scoping decision. What it does claim is that the shape of experience — its organization, its emotional tone, its sense of self — is constituted by the algorithmic structure of the models a brain runs. That claim is specific enough to generate testable predictions about complexity measures, criticality, cross-frequency coupling, and neurophenomenological correlations, and the corpus now spans enough empirical and theoretical work to make it a serious research program rather than a philosophical sketch.