Giulio Ruffini

A quantum-gravity theorist who left physics to build brains — and arrived, from inside functionalism, at consciousness as a self-modeling agent that compresses its world and selects among its own states. Mythic language and algorithmic information theory converge on the same mechanism from different ontologies.

Giulio Ruffini is a theoretical physicist who left fundamental physics to model brains, and in the process built one of the more rigorous mainstream attempts to define consciousness in mathematical terms. His Kolmogorov Theory of consciousness holds that awareness is the signature of a system that has found compressive models of its world — to be conscious is to model, and to model well is to compress. Ruffini reaches a self-modeling agent that compresses experience and selects among its own states while treating the brain as an information processor. The mechanism converges with idealist accounts while the ontology remains different.

Biography

Ruffini trained as a theoretical physicist, completing a PhD in quantum gravity before his work turned toward the practical physics of signal and measurement. His early career ran through radar and satellite reflectometry — GNSS-R, the use of reflected GPS signals to sense the Earth — the discipline of pulling structure out of noisy electromagnetic returns. The trajectory matters in the same way it does for Kastrup: he came to the mind from inside the hard sciences, trained to distinguish a real signal from an artifact, rather than through the humanities.

In 2000 he co-founded Starlab in Barcelona, a research company translating neuroscience and space science into instruments. In 2011 he co-founded Neuroelectrics, which builds non-invasive electrical brain-stimulation technology — tDCS, tACS, and related transcranial methods — driven by computational models of the cortex. In 2025 he established the Barcelona Computational Foundation, a nonprofit pursuing life, brains, and minds as fundamentally computational phenomena. Across all three the through-line is a single conviction: that the brain is a modeling engine, and that its dynamics can be read, formalized, and driven.

The Kolmogorov Theory

The theory takes its name from Kolmogorov complexity — the length of the shortest program that reproduces a given string. Ruffini’s proposal, developed across the Kolmogorov Manifesto (2007) and An algorithmic information theory of consciousness (2017), is that the brain’s central task is to discover compact generative models of the data it receives, and that structured conscious experience is what an efficient compressive model feels like from the inside. A mind is an algorithmic agent: it builds the shortest program that predicts and controls its world, and it acts to maximize an objective function that the later work grounds in affective valence. On this reading cognition is compression through and through — the search for the shortest model that predicts and controls, with nothing left over. Understanding is compression, and the felt coherence of a conscious moment is the felt coherence of a good model.

The clinical work extends this into computational neuropsychiatry: disorders of mind become disorders of the agent’s modeling and valuation — a depression is a model stuck in a bad basin — and the stimulation technology becomes a way of perturbing the dynamics back toward health. His group’s neural-mass and laminar models (the same lineage surveyed in the Rosetta Stone of Neural Mass Models) are the formal machinery underneath, the mathematics of cortical oscillation treated as the substrate of the modeling agent.

The Computational Convergence

The Human Chord describes consciousness as cross-frequency phase-locked resonance and defines agency as a bandwidth-coherence product; the neural-mass-model tradition Ruffini’s lab helped systematize is the rigorous account of how those oscillations arise and couple. The two descriptions meet at the mechanism: a conscious agent models its world through coupled oscillations and selects among possible states.

The deeper convergence is with the selection faculty. The sorting agent pulls order from noise and pays the thermodynamic cost in erasure; recursive agency measures the recurrence of self-model-mediated state selection. Ruffini’s algorithmic agent — a system that compresses its world and selects the action that best serves its objective — describes the same operation in information-theoretic language. The demon that sorts microstates and the agent that finds the shortest program are two descriptions of a system reducing the entropy of its own world-model. Sorting and compression meet at that faculty.

The ontological divide matters. Kolmogorov Theory is functionalist and near-eliminativist about the hard problem: it begins with the brain as an information processor, close to the inverse of consciousness primacy. The value lies in the independent reconstruction of the conscious faculty: two accounts disagree about what is fundamental yet converge on what the faculty does.

References

Ruffini, G. “Information, complexity, brains and reality (Kolmogorov Manifesto).” arXiv:0704.1147, 2007.

Ruffini, G. “An algorithmic information theory of consciousness.” Neuroscience of Consciousness, 2017(1), nix019.

Ruffini, G. et al. “The Algorithmic Agent Perspective and Computational Neuropsychiatry.” Entropy, 26(11), 953, 2024.

Castaldo, F., de Palma Aristides, R., Clusella, P., Garcia-Ojalvo, J., Ruffini, G. “Rosetta Stone of Neural Mass Models.” arXiv:2512.10982, 2025.

plain language The same argument in direct language. Open the full reading whenever you want the evidence, lineage, and complete architecture.

Giulio Ruffini asks a practical version of the consciousness question: what does a mind have to do in order to be aware?

His answer is compression. A mind takes an overwhelming stream of experience, finds a compact model of what matters, uses that model to predict and act, and updates it when the world resists. He begins with the brain as an information-processing system, not with consciousness as the ground of reality. Even so, his account converges with a central TimeWar concern: agency depends on a system that can make a coherent model and select among its own possible actions.

Ruffini came to mind research through physics and measurement

Ruffini trained in theoretical physics, including quantum gravity, before moving into signal processing and neuroscience. His early work involved finding useful structure in noisy reflected radio signals. That is a fitting preparation for brain research: the problem is to distinguish a real pattern from a misleading fluctuation.

He later helped found Starlab, Neuroelectrics, and the Barcelona Computational Foundation. Their work brings together computational models of the brain, brain recording, and non-invasive electrical stimulation. The shared premise is that brain activity has patterns that can be modeled and, within limits, influenced.

Consciousness, on this account, is effective world-modeling

Kolmogorov complexity is a way of asking how short the best description of something can be. A long, random-looking string needs a long description. A patterned string can be recreated by a short rule.

Ruffini applies that idea to minds. A capable mind does not store every detail separately. It finds a compact rule-set that lets it understand what is happening, anticipate what may happen next, and choose an action. The more effectively it models the world and itself, the more organized its conscious experience becomes.

This is not the claim that a person is merely a computer program. It is a functional account of what a conscious system does: it turns experience into a usable model, then uses that model to guide action. Ruffini’s later work also connects this to clinical questions. When modeling or valuation becomes stuck in a destructive pattern, the person can become stuck with it; treatment aims to help the system regain flexibility.

Compression describes an operation; it does not settle what reality is made of

The overlap with TimeWar is real. Both accounts treat agency as the capacity to find order in experience and make consequential selections rather than merely react. The sorting agent describes the same broad operation in thermodynamic language: it separates meaningful difference from noise and pays a cost for doing so. Ruffini describes it in information theory: the agent finds a compact model and chooses by it.

The difference is just as important. Ruffini starts from a physicalist or functionalist picture in which the brain performs the modeling. TimeWar holds that consciousness is fundamental and the brain is an interface within it. The two positions therefore agree about an important mechanism while disagreeing about the substrate.

That is a useful convergence, not a license to erase the disagreement. A mathematical description of how models compress and guide action does not prove consciousness is produced by the brain. It shows that any account of conscious agency has to reckon with modeling, selection, coherence, and the cost of maintaining them.

References

Read Ruffini’s Kolmogorov Manifesto and his 2017 paper, “An algorithmic information theory of consciousness,” for the original account. His later work on the algorithmic agent and neural-mass models shows how the proposal connects to computational psychiatry and brain dynamics.