Giulio Ruffini
@ruffini
Physicist working on computational neuroscience, brain stimulation and foundational aspects of (meta)physics, swimming and music during spare CPU cycles. Neuroelectrics.com, Starlab.es, BCOM.one
New preprint out! => The cortical column as a tuned receiver: a network mechanism for temporal-interference stimulation. We propose a mechanism for #TI leveraging oscillatory computation in the brain. How? Nonlinearity + resonant network amplification, like AM radio! 📻 doi.org/10.5281/zeno...
What survives the Filter of Time? Will we? What do a proton, a flame, a chemotactic cell, and an LLM agent have in common? ⚛️🧬🐕🤖 (fast intro: giulioruffini.com/blog/) @bcom-foundation.bsky.social
The things that stay. @bcom-foundation.bsky.social You cannot step into the same river twice—but the river remains. 🌊 From Aristotle to Alan Turing, how do we persist when our atoms are constantly replaced? #Philosophy #CognitiveScience #KT giulioruffini.github.io/assets/blogs...
The transformer has no memory. Some thoughts on LLMs, memory, RAG, and how to use Claude. bcomone.atlassian.net/wiki/spaces/... @bcom_foundation @bcom-foundation.bsky.social @francesca-castaldo.bsky.social
Do you need to understand the world to survive in it? A classic 1970 cybernetics theorem says yes: "Every good regulator of a system must be a model of that system." But proving this mathematically for complex, unpredictable real-world scenarios has always been notoriously difficult. 1/5
This is the paper where, after almost 20 years, I started taking my AIT obsessions a bit more seriously. I am happy I did... maybe you have some too... => Don't wast time! : ) academic.oup.com/nc/article/2... PS: the Supplementary Data part is more fun!
An Introduction to Galois Theory (with connections to AIT): zenodo.org/records/1845... ...This note aims to demystify Galois Theory by connecting its foundational definitions to a broader principle of computational and compositional tractability.
From The Sorcerer's Apprentice to Crystal Nights: Security Implications from Moltbot/Moltbook to Greg Egan's Crystal Nights! zenodo.org/records/1844...
My annotated slides for my talk in the wonderful thoughtforms.life/symposium-on... organized by @drmichaellevin are here: giulioruffini.github.io/assets/slide...
14/ Other goodies: discussion using L-operators (for synapses) and transfer functionals.
Definition: A dataset is said to represent an oscillation when it can be most succinctly Lie-generated from a representation of U1 (plus noise). #ait #kolmogorov
12/ Bonus material: plenty of good stuff in the Appendix for aficionados, including links with Groups, Topology, and Algorithmic Information Theory (What is an oscillation)? @ERC_Research @neurotwin @Neuroelectrics
8/ NMM1 (second-order synapses): PING-like motifs, Jansen–Rit, Wendling, and laminar neural masses become variants of one formalism— highlighting which parameters control resonance, PSPs, and phase shifts.
2/ Our starting point is simple: Oscillations can be seen as a push–pull interaction between two effective degrees of freedom (think E↔I, or quadrature components). That core motif survives as we add biology.
1/ 🧠 New in Computational Neuroscience: "Rosetta Stone of Neural Mass Models." An unifying framework connecting harmonic oscillator with Stuart-Landau, Wilson-Cowan, NMM1 & NMM2 (next generation). W. @Castaldo_Fr, R de Palma Aristides, P Clusella & J Garcia-Ojalvo - arxiv.org/abs/2512.109...
4/5 Proof‑of‑concept in a laminar neural mass model: a fast PING‑like circuit acts as the carrier; a slower Jansen‑Rit circuit modulates/extracts the envelope. We observe gamma sidebands and staged demodulation (Fig. 3.1, p. 14).
HAM also explains spectral “1/f” structure. In a cascade, the aperiodic slope obeys α = 2 ln(2/m)/ln r (m=modulation depth, r=spacing). A log‑uniform bank of oscillators gives a 1/f backbone (see Fig. 2.3, p. 10). 3/5
This multiplicative mixing creates intermodulation terms (fc ± Σ ai fi). To keep bands separable & demodulable, centers should be log‑spaced with ratio r ≳ 2–3 (constant‑Q)—matching canonical delta→gamma spacing (see Fig. 1.1, p. 5). 2/5
New preprint 🪢: Neural Encoding through Hierarchical Amplitude Modulation (HAM). Slower rhythms multiplicatively modulate the amplitude of faster ones, so information can live in signals, their envelopes, and envelopes‑of‑envelopes. 1/5
8) Bonus: the regulator displays an improved compression with regulator ON ⇒ the regulator shares structure with the world. Bonus: the regulator displays as-if behavior with Objective Function + Policy (algorithmic agenthood!).
7) Example (thermostat): a dead‑band bang‑bang thermostat (no integrator ⇒ fails IMP for constant refs) can still be a good algorithmic regulator if it keeps the error stream regular/compressible (e.g., a simple limit cycle).
6) Relation to IMP: The Internal Model Principle gives structural necessity (embed a copy of the exosystem for perfect regulation). Our result gives information‑theoretic necessity from single episodes, with no linearity/probability/exactness assumptions.
5) Main theorem: a statement about the posterior on the W, R pair. Meaning: a sustained compressibility advantage is evidence that the regulator contains a model of the world, formalized as positive mutual algorithmic information – M(W,R)>0.
4) Following the AIT rabbit, we score regulation by the Kolmogorov complexity of the output K(x). This is actually intuitive (e.g., perfect regulation = output is all 0s!).
News🥁! Preprint on an AIT version of the Good Regulator Theorem. World (W) and Regulator (R) are Turing machines interacting via "interface" tapes. Regulation =: compression of W output (e.g., x = temperature series of a room). #Cybernetics #AlT #ActiveInference arxiv.org/abs/2510.10300 1/9
What does Kolmogorov Complexity have to do with brain stimulation? I say ... a lot! www.dropbox.com/scl/fi/7dekr...
My slides from The Science of Consciousness (Barcelona 2025) Plenary talk this week! "From Kolmogorov Theory to Computational Modeling and Brain Stimulation". #consciousness #neuromodulation
If you like the music, consider sending a paper to our special Issue, "The Mathematics of Structured Experience: Exploring Dynamics, Topology, and Complexity in the Brain". See www.mdpi.com/journal/entr...