People often ask me where this all started. I have spent twenty years in the AI industry, and at some point the phrase "the universe is a neural network" surfaced in my head. The question is how it got there, given that I am neither a neuroscientist nor a cosmologist.
The short honest timeline is this.
2017–2019: architecture as language
I had been working with transformers and attention mechanisms since the first papers came out. I worked with systems that learned from data and demonstrated properties no one had programmed in. By 2019 I had a stable intuition: architecture matters more than algorithm. The way the connections are arranged determines what the system can and cannot do.
Engineering practice, not metaphysics. Anyone who has compared LSTM and Transformer on the same task has seen it.
2020: Vanchurin's paper
In September 2020, Vitaly Vanchurin posted a preprint on arXiv titled The World as a Neural Network. I noticed it. The title sounded like a pop-science headline — the content turned out to be a proper formalism. He showed that if you take a large neural network with the right learning dynamics, then in specific limits you can derive quantum mechanics (as the fast-learning limit) and gravity (as the slow one).
Six months later Vanchurin, with Katsnelson and Koonin, published an extension in PNAS. Koonin is a well-known evolutionary biologist; Katsnelson, a condensed-matter physicist. Serious work in a serious journal. No fringe.
It gave me an odd feeling: someone had put into words, formally, what I had felt as an engineer but could not yet say as a scientist.
2020–2022: five cracks
After that — slow accumulation. I started noticing that related observations were appearing in parallel, in different fields, with no one stitching them together.
— Vazza and Feletti (Frontiers in Physics, 2020) showed that the brain's neural network and the cosmic web are statistically indistinguishable.
— Bérut and colleagues (Nature, 2012, widely discussed 2019–2020) experimentally confirmed Landauer's principle: information has a thermodynamic cost.
— Michael Levin (Tufts) published paper after paper: cells know form before genes, bioelectric patterns encode morphology, xenobots self-organise from frog fragments into new life forms.
— Tononi sharpened IIT and began to formalise it in computationally tractable cases.
— Penrose and Hameroff got new experiments showing microtubules hold coherence longer than previously thought.
Separately — news items from five different journals. Together — the same claim stated five ways: information is the primary layer, matter and force are emergent, consciousness is a property of integrated processes.
2022–2023: the book has to be written
By 2022 I felt this synthesis had to be stated publicly in Russian — because in English it lives in fragments across academic Twitter and podcasts, and a book collecting it with an honest edge of "what is proven, what is not" had not yet been written.
I am neither a neuroscientist nor a cosmologist. But I had twenty years of work in systems where "architecture determines what is possible", and I had talked with researchers from all five fields. An unusual position: neither the depth of a specialist, nor the breadth of a journalist, but something third — an engineering perspective on science that had not yet become science.
2023–2025: Pointer Architecture
While I was writing the book, my own working hypothesis grew, and I gave it a name: Pointer Architecture. A specific information-geometric model predicting how galaxy rotation curves should behave if gravity is emergent rather than fundamental.
In 2024 I assembled the preprint, ran it against the SPARC catalogue (171 disc galaxies), and got partial confirmation with concrete preregistered falsifiers. Code and data are public.
Far from a proof of Neural Cosmology. A first measurement of one of its consequences. Roughly the way the first LHC run was an instrument for testing the Standard Model, not its proof.
Where I am now
Neural Cosmology is a working frame that is honest about its edges, not a doctrine I am promoting. Part of it is being tested, part is speculation for now, part is personal interpretation. I try not to mix those three.
If the frame is right — we have in hand the start of a new stage in science. If it is wrong — we have the first serious attempt to state and test it, which is also valuable: the scientific community gets a concrete structure to push against while searching for the correct one.
Which of the two outcomes will play out, I do not know. That is the reason I keep going.