# Neural Cosmology

Source: https://neuralcosmology.com/en

Research programme, books and essays by Mikhail Savchenko on consciousness, cosmology and computation.

- [Answers](https://neuralcosmology.com/en/answers.md)
- [Essays](https://neuralcosmology.com/en/essays.md)
- [Books](https://neuralcosmology.com/en/books.md)
- [Experiments](https://neuralcosmology.com/en/experiments.md)

Machine-readable: catalog https://neuralcosmology.com/catalog.json, full text https://neuralcosmology.com/llms-full.txt, MCP https://neuralcosmology.com/mcp.
Other languages: https://neuralcosmology.com/ru.md, https://neuralcosmology.com/pt.md, https://neuralcosmology.com/es.md


---

# Is the universe a neural network? What the evidence says

Source: https://neuralcosmology.com/en/answers/is-the-universe-a-neural-network

**Is the universe a neural network?** Nobody has shown that it is. Two lines of work make the idea serious: Vitaly Vanchurin's 2020 model, in which the learning dynamics of a network reproduce equations resembling quantum mechanics and gravity, and the 2020 Vazza–Feletti measurement showing that the brain's neuronal network and the cosmic web share statistical structure across 27 orders of magnitude. Both are hypotheses under test.

## Two different claims hide inside one question

"The universe is a neural network" can mean two things, and they need different evidence.

The **weak claim** is structural: the large-scale universe and the brain are organised in similar ways, so the same mathematics describes both. This one can be checked with a microscope, a telescope and statistics.

The **strong claim** is dynamical: the universe *learns*. Its laws are the stable result of something like training, the way the weights of a network settle after gradient descent. This one needs a formal model and predictions that differ from standard physics.

Both claims have serious work behind them. Neither is established.

## The structural evidence: brain versus cosmic web

In 2020 the astrophysicist Franco Vazza and the neurosurgeon Alberto Feletti published a quantitative comparison of two images: thin sections of human brain tissue and cosmological simulations of the cosmic web, the network of galaxy filaments and voids that fills the observable universe (*Frontiers in Physics*, 2020).

They measured the same quantities on both: how density fluctuations are distributed across scales, how many connections each node has, how clustered the network is. Several of those numbers came out close, even though the two systems differ in size by about 27 orders of magnitude. Both networks also hold most of their mass in a passive component: water in the brain, dark energy in the universe.

What this shows is that two very different physical processes arrive at comparable network statistics. Whether the cosmic web does anything like thinking is a separate question these data leave untouched. Bronchi and tree crowns share fractal statistics too, because both solve the same problem: maximise surface while keeping volume low. The interesting question the measurement leaves open is **which optimisation problem the cosmic web is solving**, if it lands on the same statistics as an evolved brain.

## The dynamical proposal: Vanchurin's model

Vitaly Vanchurin, a physicist at the University of Minnesota Duluth, made the strong claim formal in "The World as a Neural Network" (*Entropy* 22, 1210, 2020). He describes the universe as a network with two kinds of variables, trainable ones (like weights) and hidden ones (like neuron states), and studies how the system behaves while it learns.

In one limit, the learning dynamics can be written in a form resembling the Schrödinger equation. In another, the dynamics resemble Einstein's equations of general relativity. Vanchurin's argument is that quantum mechanics and gravity may be two regimes of one learning process. He later developed the idea with Yuri Wolf, Mikhail Katsnelson and Eugene Koonin into a theory of evolution as multilevel learning (*PNAS*, 2022).

The model is a proposal. It has not produced a measurement that only it can explain.

## A third thread: gravity from information

In 2010 Erik Verlinde argued that gravity may be emergent, a consequence of how information is distributed on surfaces, and in 2017 he extended the argument to galaxy dynamics. If gravity comes from information, and a network is a machine for moving information around, then the two ideas meet: the pattern of gravity would be the connection pattern of the network. That is the step this programme takes.

## What Neural Cosmology adds: a working substrate

The programme on this site takes the dynamical claim one level lower and asks what the smallest network capable of producing physics-like structure would look like.

Its preprint, *Pointer Architecture* v9.0, defines a formal substrate: pointers, rewrite rules, commits, an append-only archive and observers. It implements that substrate as a small programming language, Sixth. Forty demonstrations, checked by 646 automated assertions, build up from a single distinction to arithmetic, time, space, observers and universal computation, and on to self-maintaining structures. On the same substrate, a holographic construction reproduces the standard estimate of dark energy in substrate language, within a factor of about 0.73 of the observed value.

What this establishes is constructive: a network of this kind can generate the structures a network universe would need. Whether our universe runs on one is a separate question, and the paper states in advance what would break its claims.

## What would count against the idea

- A substrate demonstration failing its automated checks on the released code (falsifier F0 in the preprint).
- Area scaling breaking on other substrate topologies, which would make the dark-energy result an artefact of the chosen lattice (F1).
- Brain–cosmic web similarity disappearing under better statistics or appearing equally in random networks with the same density.

## Bottom line

The universe being a neural network is a live scientific hypothesis with formal models, one striking structural comparison and a first working substrate built as code. It has not yet become a finding. It should be judged the way any model is judged: by the predictions it makes and by what could break it. The book [The Celestial Code](https://neuralcosmology.com/en/books/celestial-code) walks through the full argument in plain language.

## FAQ

### Who proposed that the universe is a neural network?

The best-known formal proposal is Vitaly Vanchurin's paper “The World as a Neural Network” (Entropy, 2020). He treats the universe as a network of trainable and hidden variables and shows that, in different limits, its learning dynamics look like quantum mechanics and like general relativity.

### Does the cosmic web really look like the brain?

Statistically, in some respects, yes. Franco Vazza and Alberto Feletti (Frontiers in Physics, 2020) compared brain tissue with simulations of the cosmic web and found similar fluctuation spectra and similar numbers of connections per node, despite a difference in scale of about 27 orders of magnitude. Similar statistics do not mean the two systems do the same thing.

### Is this the same as the simulation hypothesis?

No. The simulation hypothesis needs a simulator outside the universe. A network universe needs no programmer: the claim is that physical law itself behaves like a learning process.

### How could the idea be tested?

By building a concrete network model, deriving what it predicts and stating in advance what would falsify it. The Pointer Architecture preprint implements such a substrate as working code, the Sixth language, and lists its falsifiers in the manuscript.


---

# Galaxy rotation curves without dark matter: MOND, emergent gravity and what the data say

Source: https://neuralcosmology.com/en/answers/galaxy-rotation-curves-without-dark-matter

**Can galaxy rotation curves be explained without dark matter?** Partly. MOND fits many individual rotation curves with a single acceleration scale, and emergent-gravity models reproduce the radial acceleration relation, but neither explains galaxy clusters and the cosmic microwave background as well as dark matter does.

## The problem in one picture

Measure how fast stars orbit at different distances from a galaxy's centre and you get a rotation curve. Newtonian gravity applied to the matter we can see predicts that the curve should rise and then fall, the way outer planets move more slowly than inner ones. Real spiral galaxies do not do that. Their curves rise and then stay flat far beyond the visible disk.

Vera Rubin and Kent Ford showed this for Andromeda in 1970 (*Astrophysical Journal* 159, 379), and a decade of measurements made it the rule. Something adds gravity where there is little visible matter.

## Answer 1: dark matter halos

The standard answer is invisible matter distributed in a roughly spherical halo around each galaxy. Cosmological simulations predict halo shapes, the best known being the NFW profile of Navarro, Frenk and White (1996). With a halo fitted to each galaxy, rotation curves are reproduced well.

Dark matter also explains things far from galaxies: the pattern of peaks in the cosmic microwave background, the growth of large-scale structure and the Bullet Cluster, where the lensing mass sits apart from the colliding gas. That is why most cosmologists treat it as the default.

Its weak points are at galactic scales. Simulated halos tend to be too dense in the centre (the core–cusp problem), and each galaxy needs its own halo parameters. And after decades of searches, no dark matter particle has been detected in the lab.

## Answer 2: MOND

In 1983 Mordehai Milgrom proposed changing the law instead of adding matter. Below an acceleration of about 1.2 × 10⁻¹⁰ m/s², gravity would weaken more slowly than Newton says. With that one constant, MOND predicts many rotation curves from visible matter alone.

McGaugh, Lelli and Schombert (2016) found a tight **radial acceleration relation** in SPARC galaxies: the observed acceleration is a simple function of the acceleration from visible matter. MOND predicted this. For dark matter it is a coincidence that has to be explained.

MOND's weak points are the reverse of dark matter's: galaxy clusters still need extra mass, and relativistic versions have trouble with the cosmic microwave background.

## Answer 3: emergent gravity

Erik Verlinde argued in 2010 that gravity is not fundamental and emerges from how information is stored on surfaces. In 2017 he derived from this an extra gravitational effect on galactic scales that mimics dark matter and reproduces a MOND-like acceleration scale. The model is young, and its tests on real galaxies are mixed.

## Where this programme stands

The research programme behind this site does not offer its own rotation-curve model. Its preprint, *Pointer Architecture* v9.0, works at the level of a formal computational substrate. Its one cosmological statement is a substrate-language reproduction of the standard holographic estimate of dark energy, which concerns the expansion of the universe and says nothing specific about the dynamics of individual galaxies.

## Bottom line

Rotation curves can be fitted without dark matter at galactic scales, and a simple law (MOND) predicts them surprisingly well. No alternative yet matches dark matter across clusters and cosmology. The sober position is that the problem is open at the scale of galaxies, and the way forward is models that make falsifiable predictions on public data.

## FAQ

### What is a galaxy rotation curve?

A plot of how fast stars and gas orbit a galaxy's centre as a function of distance. If visible matter were all there is, speeds far from the centre should fall off like planets around the Sun. Observed curves stay roughly flat instead.

### Who discovered flat rotation curves?

Vera Rubin and Kent Ford measured the rotation of the Andromeda galaxy in 1970, and through the 1970s Rubin's group showed that flat curves are the rule for spiral galaxies. Fritz Zwicky had noticed missing mass in the Coma cluster back in 1933.

### What is MOND?

Modified Newtonian Dynamics, proposed by Mordehai Milgrom in 1983. It changes Newton's law below an acceleration of about 1.2 × 10⁻¹⁰ m/s², which reproduces many rotation curves without dark matter. It struggles with galaxy clusters and cosmology.

### What is the SPARC database?

Spitzer Photometry and Accurate Rotation Curves: 175 disk galaxies with near-infrared photometry and high-quality rotation curves, published by Lelli, McGaugh and Schombert in 2016. It is the standard public sample for testing rotation-curve models.

### Does the Bullet Cluster rule out alternatives to dark matter?

It is strong evidence for some form of unseen mass, because gravitational lensing and hot gas are displaced from each other after the cluster collision. Alternatives have to reproduce it, and most have difficulty doing so.


---

# Michael Levin and bioelectricity: how cells store the plan of the body

Source: https://neuralcosmology.com/en/answers/michael-levin-bioelectricity

**What did Michael Levin discover about bioelectricity?** Michael Levin's lab at Tufts showed that patterns of voltage across cell membranes carry instructions for body shape. Changing those patterns made flatworms grow two heads, and the two-headed form persisted through later regeneration with unchanged DNA. Frog embryos grew eyes on their tails. Part of anatomical memory is stored in bioelectric networks, outside the genome.

## The experiment that changed the question

Planarian flatworms regenerate: cut one in half and each half grows back what is missing. That has been known since the nineteenth century.

Levin's lab changed the membrane voltage pattern in regenerating fragments using drugs that act on ion channels. The fragments grew a second head where the tail should have been. The surprising part came next. When the two-headed worms were cut again, without any further treatment, the pieces regenerated **two heads again**, over several rounds, with genomes identical to normal worms (Durant et al., *Biophysical Journal*, 2017).

If the DNA is unchanged and every tissue has been rebuilt, the instruction "build two heads" has to be stored somewhere else. The candidate is the bioelectric pattern itself: a stable distribution of voltages across a network of connected cells.

## Eyes on tails, faces in voltage

In frog embryos (*Xenopus*), Pai and colleagues imposed a specific voltage state on cells far from the head and induced eyes to form there, including on the tail and gut (*Development* 139, 313, 2012). Earlier, the lab imaged a voltage "pre-pattern" of the face appearing across the embryo before the face's anatomy develops.

The voltage map seems to act like a set of instructions or a target that tissue works towards.

## Xenobots and anthrobots

In 2020 Kriegman, Blackiston, Levin and Bongard described **xenobots** (*PNAS* 117, 1853): clusters of frog skin and heart cells, arranged by a design algorithm, that move, push particles together and heal. Their genome was never edited. Later work from the lab built **anthrobots** from adult human airway cells, which crawl and can encourage neurons to grow across a scratch in a dish.

The point for theory: cells whose developmental program ends in skin can coordinate into new, working body forms that no genome encoded.

## Levin's framework: cognition as a continuum

Levin interprets these results in terms of goals and memory at many scales. In his TAME framework (Technological Approach to Mind Everywhere, 2022), a cell collective counts as an agent to the degree that it pursues a target state despite perturbations. Its "cognitive light cone" is the size of the goals it can pursue.

This is a working vocabulary with experimental payoffs, which is why it spread quickly. It stays agnostic about subjective experience.

## Why this matters for Neural Cosmology

The research programme on this site proposes that memory and form can live in a shared, addressable structure that local units read from, a scheme it calls Pointer Architecture. Levin's results are the clearest biological case of that scheme: cells behave like clients reading a stored target, and the target survives the replacement of the cells.

That view makes a prediction Levin's methods could test: **a bioelectric pattern should be transferable** between organisms without transferring genetic material. Impose the voltage landscape of a two-headed planarian on a normal one and it should start building two heads at its next regeneration. If that fails under good conditions, this leg of the programme fails too. The full reasoning is in the essay *Levin's hologram*.

## What remains unknown

- How a voltage pattern encodes a specific anatomy, and at what resolution.
- Whether the stored pattern is local to the tissue or depends on long-range coordination.
- Where information processing in cell collectives ends and anything like experience would begin. Nobody has a measurement for that yet.

## Bottom line

Levin showed that bioelectric networks store and execute part of the body plan, independently of changes to DNA. That makes biology one of the best places to test ideas about distributed memory, including the ideas behind this programme.

## FAQ

### Who is Michael Levin?

A developmental and synthetic biologist at Tufts University, director of the Allen Discovery Center. His lab studies how cells use bioelectric signals to coordinate growth, regeneration and body shape.

### What is bioelectricity in biology?

The voltage differences across cell membranes created by ion channels and pumps, and the networks through which cells share those voltages via gap junctions. Besides nerve impulses, these patterns guide how tissues grow and where organs form.

### What are xenobots?

Small living constructs made from frog (Xenopus) skin and heart-muscle cells, designed with an evolutionary algorithm and described in PNAS in 2020 by Kriegman, Blackiston, Levin and Bongard. With an unchanged genome they move, gather particles and repair themselves.

### Does Levin's work prove that cells are conscious?

No. It shows that cell collectives process information and pursue anatomical goals. Levin argues for a continuum of cognition, but whether any of this involves experience is a separate, open question.


---

# The hard problem of consciousness: the main theories and what can be tested

Source: https://neuralcosmology.com/en/answers/hard-problem-of-consciousness

**What is the hard problem of consciousness, and which theories try to solve it?** The hard problem, named by David Chalmers in 1995, asks why physical processes in the brain are accompanied by subjective experience at all. The leading theories (global workspace, integrated information, higher-order and predictive processing) mostly explain which processes are conscious. None explains why they feel like anything. The useful test is what each theory predicts that the others do not.

## Easy problems and the hard one

In 1995 the philosopher David Chalmers split the question of consciousness in two (*Journal of Consciousness Studies* 2, 200).

The **easy problems** are about functions: how the brain tells red from green, integrates signals, focuses attention and reports "I see red". They are difficult, and neuroscience makes steady progress on them.

The **hard problem** is why any of this is accompanied by experience, why there is *something it is like* to see red. You could, in principle, explain every function and still not have explained that.

Some philosophers (Daniel Dennett, Keith Frankish) argue that the hard problem dissolves once the functions are explained, and that the sense of an extra ingredient is an illusion the brain produces. Others think it points to a gap in physics itself. That disagreement is still open.

## The main theories, briefly

**Global workspace theory** (Bernard Baars, 1988; Stanislas Dehaene and colleagues). Information becomes conscious when it wins a competition and is broadcast across a brain-wide network, so that many systems can use it at once. It predicts a sudden "ignition" of frontal and parietal activity when a stimulus becomes conscious.

**Integrated information theory, IIT** (Giulio Tononi, 2004). Consciousness is integrated information, measured by a quantity Φ: how much a system as a whole constrains its own past and future beyond its parts. It predicts that consciousness depends on the causal structure of posterior cortex more than on frontal broadcast. A practical offshoot, the perturbational complexity index (Casali et al., 2013), distinguishes wakefulness, sleep, anaesthesia and disorders of consciousness from EEG responses to magnetic pulses.

**Higher-order theories** (David Rosenthal and others). A state is conscious when the brain represents itself as being in that state.

**Predictive processing** (Andy Clark, Anil Seth, Karl Friston). The brain is a prediction machine; experience is the content of its best guesses about the world and the body.

**Orch-OR** (Roger Penrose and Stuart Hameroff, 1996). Consciousness arises from quantum processes in neuronal microtubules. It is the most physics-heavy proposal and the most disputed.

## What the head-to-head tests found

A consortium of labs (Cogitate) ran preregistered experiments designed so that GWT and IIT would predict different results. The findings, released in 2023 and published in *Nature* in 2025, supported some predictions of each theory and contradicted others: the sustained synchronisation within posterior cortex that IIT predicted did not appear, and neither did the "ignition" at stimulus offset that GWT predicted.

The lasting lesson is methodological. Theories of consciousness are now judged by preregistered predictions that can fail. That shift matters more than any single result.

## Where the information-architecture view fits

The research programme on this site treats consciousness as a problem about **how information is addressed and read**, not only how much of it is integrated. In that picture, a conscious system is one that reads from and writes to a shared structure beyond its own parts, a pattern the programme calls Pointer Architecture and implements as working code (the Sixth language) with a substrate-side consciousness measure, Φ_PA.

For consciousness, the view makes one risky prediction, discussed in the essay *Attention as load*: if attention is a read operation on a shared structure, mass focused attention should have a small, measurable effect on physical random processes under a preregistered protocol. Earlier experiments of this kind (PEAR, the Global Consciousness Project) produced small, contested effects. A clean null result under a strict protocol would count against the idea.

## Bottom line

Science has good and improving theories of which brain processes are conscious. None has closed the hard problem. Theories now compete on testable predictions, and any serious new proposal, including this programme's, has to be judged the same way.

## FAQ

### What are the easy problems of consciousness?

Explaining functions: how the brain discriminates stimuli, integrates information, reports its states, controls attention and behaviour. They are hard in practice, but they are problems about mechanism, and neuroscience knows how to attack them.

### Which theory of consciousness is most accepted?

None has consensus. Global workspace theory and integrated information theory have the most experimental programmes behind them; higher-order and predictive-processing theories are also influential. Head-to-head tests published since 2023 found partial support and clear failures for both GWT and IIT.

### Is IIT pseudoscience?

In 2023 more than a hundred researchers signed an open letter calling it pseudoscience, and many others objected to that label. IIT does make some testable predictions, such as the perturbational complexity index used clinically, while its core quantity Φ cannot be computed for real brains.

### Is consciousness quantum?

There is no accepted evidence that it is. The Penrose–Hameroff Orch-OR theory proposes a quantum role for microtubules; it remains controversial and is actively being tested.


---

# Panpsychism explained: is consciousness a basic feature of matter?

Source: https://neuralcosmology.com/en/answers/panpsychism

**What is panpsychism, and is the universe conscious?** Panpsychism is the view that experience is a fundamental feature of matter, present in a tiny form in basic physical entities, with brains combining it into rich minds. Philosophers such as Chalmers, Goff and Strawson take it seriously because it sidesteps the hard problem. Its main difficulties are the combination problem and the lack of testable predictions. No evidence shows that the universe as a whole is conscious.

## The claim

Panpsychism says that experience does not appear out of nothing when matter gets organised into a brain. It is there from the start, as a basic feature of the physical world, in some minimal form. Brains are special because they combine and concentrate it.

Put that way it sounds mystical. The modern argument for it is dry and comes from the hard problem.

## Why philosophers take it seriously

Physics describes what matter *does*: masses, charges and how they interact. It says nothing about what matter *is* intrinsically. This observation goes back to Bertrand Russell and Arthur Eddington in the 1920s.

We know one case where we have access to the intrinsic nature of some matter: our own brains, from the inside, as experience. **Russellian monism** proposes the simplest fit: the intrinsic nature that physics leaves blank is experiential, or proto-experiential, everywhere. Then consciousness has a place in nature without adding anything to the equations.

David Chalmers, Philip Goff (*Galileo's Error*, 2019) and Galen Strawson have developed versions of this argument. It explains why panpsychism moved from fringe to a respectable position in philosophy of mind over the last twenty years.

## The problems

**Combination.** If an electron has a flicker of experience, how do the flickers of a hundred billion neurons' worth of particles become one unified point of view? Adding small experiences does not obviously produce a single larger subject. This is the problem panpsychists themselves consider the hardest.

**Testability.** Panpsychism, as usually stated, predicts no observation that a physicalist would not also predict. A stone behaves the same whether or not it has a trace of experience. A hypothesis that forbids nothing is hard to count as science.

**Where the line goes.** If everything has experience, the interesting question becomes which systems have *unified* experience, and that needs a structural criterion, which brings us back to theories like IIT.

## Relatives: IIT and cosmopsychism

Integrated information theory is sometimes called panpsychist because it attributes some consciousness to any system with integrated information. It is more selective than classic panpsychism: a pile of sand has almost none, however many particles it contains.

**Cosmopsychism** flips the direction: the universe is the basic subject, and individual minds are fragments of it. It avoids the combination problem and inherits a mirror-image one: how does one cosmic subject split into many private viewpoints?

## Is the universe conscious?

There is no evidence that it is, and no agreed way to measure it. What can be measured is structure: how information is integrated and exchanged at different scales. The brain and the cosmic web share some network statistics, and physical law can be written in informational terms. Those facts make the question of consciousness at cosmic scale thinkable. They do not answer it.

The research programme on this site stays on the measurable side. It asks whether the universe *processes information like a network* and builds that idea as a working computational substrate with falsifiers stated in advance. Whether such processing is accompanied by experience is left open until someone finds a way to measure it.

## Bottom line

Panpsychism is a serious philosophical answer to the hard problem with two unsolved difficulties: combination and testability. A version worth adopting in science would have to name a structural quantity and predict where rich experience appears and where it does not. IIT is the closest attempt so far.

## FAQ

### Who are the main panpsychist philosophers?

Contemporary defenders include Philip Goff (Galileo's Error, 2019), Galen Strawson, and Hedda Hassel Mørch; David Chalmers treats it as one of the most promising options. Historically, versions appear in Spinoza, Leibniz, William James and Alfred North Whitehead.

### What is the combination problem?

If electrons have tiny experiences, how do billions of them combine into one unified human experience instead of billions of separate ones? No panpsychist account of combination is widely accepted.

### What is cosmopsychism?

The reverse move: the universe as a whole is the basic conscious subject, and individual minds are parts or aspects of it. It trades the combination problem for a decomposition problem.

### Is integrated information theory panpsychist?

Partly. IIT says any system with integrated information above zero has some experience, so it attributes consciousness far beyond brains. But it denies consciousness to aggregates like rocks, whose Φ is near zero, which classic panpsychism does not.


---

# Simulation theory: are we living in a simulation, or is the universe computing itself?

Source: https://neuralcosmology.com/en/answers/is-the-universe-a-simulation

**Are we living in a simulation?** There is no evidence that we are. Nick Bostrom's 2003 argument shows only that one of three possibilities must hold, and most versions of the simulation hypothesis make no testable predictions. A more productive idea is that the universe computes without an outside programmer: physical law behaves like information processing. That version can be tested against data.

## Bostrom's argument in plain terms

Nick Bostrom's paper "Are You Living in a Computer Simulation?" (*Philosophical Quarterly*, 2003) makes a narrower claim than its reputation: one of three statements is almost certainly true.

1. Nearly all civilisations at our stage go extinct before they can run realistic simulations of minds.
2. Nearly all civilisations that can run them choose not to.
3. We are almost certainly living in a simulation, because simulated minds would vastly outnumber original ones.

Which one is true depends on numbers nobody knows. The argument is logically neat and empirically empty until those numbers can be estimated.

## Why the hypothesis is hard to test

A perfect simulation would reproduce physics exactly, so no experiment inside it could tell the difference. Any test has to assume an imperfect simulation with some specific shortcut.

The best-known proposal is by Silas Beane, Zohreh Davoudi and Martin Savage (2012). If the simulation runs on a space-time grid, like today's lattice calculations of quantum chromodynamics, the grid would impose a maximum energy and a preferred set of directions. The highest-energy cosmic rays would then arrive in a slightly non-uniform pattern tied to the grid's axes. No such pattern has been seen. That rules out only this particular kind of simulation.

Other suggested "glitches", from quantum measurement to the speed of light as a processing limit, are analogies. They explain known physics after the fact without predicting anything new.

## Simulated versus computational

The simulation hypothesis bundles two different ideas:

- **The universe is computational.** Its laws describe how information is transformed. This needs no outside computer.
- **Someone is running it.** There is a simulator, a programmer and a purpose. This adds an agent for which we have no evidence.

The first idea has a long scientific history: Konrad Zuse's *Calculating Space* (1969), John Wheeler's "it from bit" (1989), Rolf Landauer's argument that information is physical, Seth Lloyd's view of the universe as a quantum computer, and Stephen Wolfram's hypergraph models (2020). It can be turned into mathematics, and mathematics can be tested.

The second idea is what makes the simulation hypothesis popular and what makes it untestable.

## The testable version

The research programme on this site takes the first idea and drops the second. It treats the universe as a network that processes information, closer to Vanchurin's world-as-a-neural-network model than to a video game, and asks what that predicts for something measurable.

Its preprint builds such a network as working code, the Sixth language, and derives from it complexity results, correspondences with known models of distributed computation and a substrate-language version of the standard dark-energy estimate, each with falsifiers stated in advance.

The novel [Bugs Academy](https://neuralcosmology.com/en/books/bugs-academy) plays the simulation idea out as fiction: a debugger who treats physical anomalies as bugs. The non-fiction [The Celestial Code](https://neuralcosmology.com/en/books/celestial-code) keeps to what can be measured.

## Bottom line

Nobody knows whether we are in a simulation, and most forms of the idea cannot be tested. The part worth keeping is the claim that physics is information processing, because it leads to models that can be checked against data.

## FAQ

### What is Bostrom's simulation argument?

In 2003 Nick Bostrom argued that at least one is true: almost no civilisations reach the ability to run detailed simulations of their ancestors; almost none that can, choose to; or we are almost certainly living in such a simulation. It is an argument about probabilities, not evidence that we are simulated.

### Can the simulation hypothesis be tested?

Only specific versions. Beane, Davoudi and Savage (2012) showed that a universe simulated on a space-time lattice could leave a signature in the angular distribution of the highest-energy cosmic rays. No such signature has been found.

### What is digital physics?

The family of ideas that the universe is fundamentally computational: Konrad Zuse's Calculating Space (1969), John Wheeler's “it from bit” (1989), Seth Lloyd's quantum-computing universe and Stephen Wolfram's hypergraph models (2020).

### Is the Neural Cosmology programme a simulation theory?

No. It does not assume a simulator. It proposes that the universe processes information like a network and builds that idea as working code with falsifiers stated in advance.


---

# Landauer's principle: why erasing information costs energy

Source: https://neuralcosmology.com/en/answers/landauer-principle

**What is Landauer's principle?** Landauer's principle, stated by Rolf Landauer in 1961, says that erasing one bit of information must release at least kT ln 2 of heat, about 2.9 × 10⁻²¹ joules at room temperature. It was measured directly in 2012. It connects information to thermodynamics and is the strongest argument that information is a physical quantity.

## The statement

In 1961 Rolf Landauer, a physicist at IBM, asked what the minimum energy cost of computation is (*IBM Journal of Research and Development* 5, 183). His answer: most operations can in principle be done without any cost, but **erasing** information cannot.

Erasing one bit, resetting a memory cell to 0 regardless of whether it held 0 or 1, must release at least

**E = kT ln 2**

of heat into the environment, where k is Boltzmann's constant and T is the temperature. At room temperature (300 K) that is about **2.9 × 10⁻²¹ J**, roughly 0.018 electron-volts.

## Why erasure has a cost

A memory that can hold 0 or 1 has two possible states. After erasure it has one. The number of possible states of the memory has halved, so its entropy has dropped by k ln 2. The second law of thermodynamics does not allow total entropy to fall, so at least that much entropy has to go somewhere else, as heat into the surroundings.

The key step is that the bit is a physical state of a physical system. Landauer summed it up in a phrase that became a slogan: **information is physical.**

## Maxwell's demon, resolved

In 1867 James Clerk Maxwell imagined a tiny being that opens a door between two gas chambers to let fast molecules one way and slow ones the other. It would create a temperature difference, apparently for free, against the second law.

Charles Bennett showed in 1982 where the cost hides. To sort molecules the demon has to record what it saw. To keep working it must eventually erase those records, and by Landauer's principle the erasure dissipates at least as much entropy as the sorting removed. The second law survives because memory is physical.

## The measurement

For fifty years the principle was a theoretical result. In 2012 Antoine Bérut and colleagues tested it directly (*Nature* 483, 187). They trapped a single glass bead in a double-well potential made by lasers, where "left" and "right" served as 0 and 1, and performed erasure cycles while measuring the heat. As the cycles were run more slowly, the released heat approached kT ln 2 and did not go below it. Experiments with nanomagnetic bits (Hong et al., *Science Advances*, 2016) and quantum systems followed.

## What it means for computers

Today's chips dissipate many orders of magnitude more energy per operation than the Landauer limit. The limit matters for the future: it sets a floor for irreversible computing and motivates reversible computing, where logic is designed so that no information is erased and the floor can, in principle, be avoided.

## Why it matters for physics built on information

Landauer's principle is the best-established link between information and energy. It is one of the foundations for ideas that treat information as a primary physical quantity: Wheeler's "it from bit", Bekenstein's entropy of black holes, Verlinde's emergent gravity.

The research programme on this site builds the same link into its substrate. In the *Pointer Architecture* preprint, history is kept in an append-only archive that is never erased, and the paper proves that this archive can be compressed at the Shannon rate. In Landauer's terms, the substrate never pays the cost of erasure, because nothing is erased; its records are compressed instead.

## Bottom line

Erasing a bit costs at least kT ln 2 of heat. The principle is experimentally confirmed, it resolved Maxwell's demon and it makes "information is physical" a measurable statement.

## FAQ

### What is the Landauer limit in numbers?

kT ln 2 per erased bit: about 2.9 × 10⁻²¹ J, or 0.018 eV, at 300 K. Erasing a billion bits at that limit would release about 3 × 10⁻¹² J.

### Has Landauer's principle been confirmed experimentally?

Yes. Bérut and colleagues measured the heat released when erasing a one-bit memory made of a colloidal bead in a double-well laser trap and found it approaches kT ln 2 (Nature, 2012). Later experiments confirmed it in nanomagnets and quantum systems.

### How far are computers from the Landauer limit?

Modern transistors dissipate far more energy per logical operation than kT ln 2, by several orders of magnitude, so the limit is not what constrains chips today. It becomes relevant for ultra-low-power and reversible computing.

### What does Landauer's principle have to do with Maxwell's demon?

The demon seemed to lower entropy for free by sorting molecules. Charles Bennett showed in 1982 that the demon must eventually erase its memory of the molecules, and by Landauer's principle that erasure produces at least as much entropy as the sorting removed.


---

# What is the cosmic web? Filaments, voids and the largest structure in the universe

Source: https://neuralcosmology.com/en/answers/cosmic-web

**What is the cosmic web?** The cosmic web is the large-scale structure of the universe: galaxies and dark matter arranged in long filaments and flat walls that meet at dense clusters, surrounding enormous, nearly empty voids. It grew from tiny density ripples in the early universe under gravity. Most galaxies sit in filaments and walls, while voids fill most of the volume.

## A map on the largest scale

On the scale of a galaxy, matter looks clumpy and local. Zoom out by a factor of a thousand and a pattern appears: galaxies line up along filaments and sheets, which meet in dense knots and wrap around huge empty regions. Astronomers call this pattern the **cosmic web**.

Its parts have names:

- **Nodes**: galaxy clusters, the densest places, where filaments intersect.
- **Filaments**: long bridges of galaxies, gas and dark matter connecting the nodes.
- **Walls** (sheets): flattened structures between filaments.
- **Voids**: nearly empty regions that take up most of the volume of the universe.

## How it formed

The early universe was almost perfectly smooth, with density ripples of about one part in a hundred thousand, the same ripples seen in the cosmic microwave background. Gravity amplified them. Slightly denser regions pulled in matter, emptier regions emptied further.

In 1970 Yakov Zeldovich showed that collapse under gravity happens first along one axis, producing flattened "pancakes", then filaments, then knots. That sequence, sheets to filaments to clusters, is still how the web is described. Dark matter does most of the pulling because there is about five times more of it than ordinary matter; gas and galaxies settle into the dark-matter scaffolding.

Large computer simulations such as the Millennium Simulation (Springel et al., *Nature*, 2005) reproduce the web from these initial ripples very closely. The match between simulated and observed structure is one of the strongest supports for the standard cosmological model.

## How it is mapped

The first clear sign came from redshift surveys in the 1980s. In 1989 Margaret Geller and John Huchra found the "Great Wall", a sheet of galaxies hundreds of millions of light-years long. Later surveys, above all the Sloan Digital Sky Survey, mapped millions of galaxies and turned the web into a detailed three-dimensional atlas.

In 2014 Brent Tully and colleagues used galaxy motions to define our home supercluster, **Laniakea**, about 500 million light-years across, with the Milky Way on its outskirts (*Nature*, 2014).

Seeing the gas in filaments directly is harder because it is thin. Since 2019 deep observations have imaged glowing gas along individual filaments, and in 2020 fast radio bursts, millisecond flashes whose signal is slowed by free electrons along the way, were used to count the diffuse ordinary matter between galaxies (Macquart et al., *Nature*, 2020). The result filled much of the long-standing gap in the census of ordinary matter.

## The brain comparison

Pictures of the cosmic web and of neural tissue look alike, and in 2020 Franco Vazza and Alberto Feletti checked whether the resemblance survives measurement. For several statistics, such as how density fluctuations distribute across scales and how many connections each node has, the numbers came out close, despite a difference in size of about 27 orders of magnitude.

The careful reading: two very different processes arrive at comparable network statistics, probably because both solve a similar problem, connecting many nodes with little material. The question of what that shared problem is, and whether the universe processes information like a network, is taken up on the page [Is the universe a neural network?](https://neuralcosmology.com/en/answers/is-the-universe-a-neural-network).

## Open questions

- How galaxies inherit their spin and shape from the filaments they sit in.
- How much of ordinary matter sits in the hottest, thinnest gas of filaments.
- Whether voids are as empty and as numerous as the standard model predicts; void statistics are becoming a test of dark energy and gravity.

## Bottom line

The cosmic web is the skeleton of the universe: filaments, walls and clusters around vast voids, grown by gravity from tiny early ripples and dominated by dark matter. It is mapped in three dimensions by galaxy surveys and, increasingly, seen directly in the gas of its filaments.

## FAQ

### How big is the cosmic web?

It spans the whole observable universe. Individual filaments can be tens to hundreds of millions of light-years long, and voids are typically tens to hundreds of millions of light-years across. Our own galaxy sits in the Laniakea supercluster, mapped in 2014, which is about 500 million light-years across.

### What is the cosmic web made of?

Mostly dark matter, which forms the scaffolding, plus ordinary gas and galaxies that follow it. A large share of ordinary matter is thin, hot gas inside filaments, which is why it long went missing from galaxy counts.

### Has anyone seen the cosmic web directly?

Mostly it is mapped from galaxy positions in redshift surveys. Since the late 2010s telescopes have also imaged the faint glow of gas in individual filaments, and fast radio bursts have been used to weigh the diffuse gas between galaxies.

### Does the cosmic web look like a brain?

In some statistical measures, yes. A 2020 study by Vazza and Feletti found similar fluctuation spectra and connectivity in brain tissue and cosmic-web simulations. Similar statistics suggest that both networks solve comparable optimisation problems; whether the universe thinks is a separate question.


---

# Is consciousness quantum? Orch-OR, decoherence and what experiments show

Source: https://neuralcosmology.com/en/answers/is-consciousness-quantum

**Is consciousness quantum?** There is no accepted evidence that it is. The Penrose–Hameroff Orch-OR theory proposes that consciousness arises from quantum processes in neuronal microtubules, but Max Tegmark's 2000 estimate suggests such states would decohere far too fast to matter. Recent anaesthesia and microtubule experiments keep the question open without settling it.

## Why people link consciousness and quantum physics

The link has two sources. One is the measurement problem: in quantum mechanics, a system seems to have definite properties only when measured, and early physicists such as John von Neumann and Eugene Wigner wondered whether the observer's mind plays a role. The other is the hard problem of consciousness: if ordinary neuroscience cannot explain why experience exists, perhaps a deeper layer of physics can.

Most physicists today separate the two. Measurement does not require a conscious observer; any interaction with the environment will do. But the second motivation keeps quantum theories of mind alive.

## The main proposal: Orch-OR

Roger Penrose argued in *The Emperor's New Mind* (1989) that human understanding is not an algorithm, and therefore needs physics beyond standard computation. With the anaesthesiologist Stuart Hameroff he developed **Orchestrated Objective Reduction** (Orch-OR): quantum superpositions form in microtubules, protein tubes inside neurons, and collapse when a gravity-related threshold is reached. Each orchestrated collapse would be a moment of experience.

The theory makes contact with data: general anaesthetics bind to many proteins, including tubulin, the building block of microtubules, which fits Hameroff's view that anaesthesia works by blocking quantum processes there.

## The decoherence objection

Quantum superpositions survive only while isolated from their environment. In 2000 Max Tegmark estimated that in the warm, wet brain, superpositions in neurons or microtubules would last roughly 10⁻¹³ to 10⁻²⁰ seconds (*Physical Review E* 61, 4194). Neural processing runs on the scale of milliseconds, a gap of ten or more orders of magnitude. Hameroff and colleagues replied that Tegmark modelled the wrong states and ignored shielding; the dispute continues.

## What experiments show

- **Quantum biology is real in places.** Enzymes use tunnelling, and bird magnetoreception most likely relies on quantum spin states in cryptochrome proteins. These show that biology can exploit quantum effects; whether the brain does so for consciousness is a separate question.
- **Microtubules and anaesthesia.** In a 2024 rat study (*eNeuro*), a drug that stabilises microtubules delayed loss of consciousness under the anaesthetic isoflurane. It is consistent with Orch-OR's prediction, and also with less exotic explanations.
- **Gravity-related collapse.** A 2020 experiment in the Gran Sasso underground laboratory searched for faint radiation predicted by the simplest Diósi–Penrose collapse model and found none (*Nature Physics*, 2021), ruling that version out. Orch-OR relies on a related but not identical mechanism.
- **Entanglement in the brain.** A 2022 MRI study claimed a signal of entanglement in the brain; it has not been independently confirmed and its interpretation is contested.

## What would decide it

A convincing case would need a quantum effect that lasts long enough to influence neural firing, measured in living brain tissue, and shown to track conscious states rather than just anaesthesia chemistry. No experiment has done that yet.

The research programme on this site does not depend on quantum effects in the brain. It treats consciousness as a question about how information is addressed and read in a network, and its preprint defines a consciousness measure, Φ_PA, on a classical computational substrate. See [the hard problem of consciousness](https://neuralcosmology.com/en/answers/hard-problem-of-consciousness) for how the main theories compare.

## Bottom line

Consciousness may or may not involve quantum effects; the evidence does not currently require them. Orch-OR is the most developed quantum proposal and has become more testable, but decoherence remains its main obstacle.

## FAQ

### What is Orch-OR?

Orchestrated Objective Reduction, a theory by Roger Penrose and Stuart Hameroff (1990s). It proposes that quantum superpositions in microtubules inside neurons collapse through a gravity-related process, and that each collapse is a moment of conscious experience.

### Why do many physicists doubt it?

The brain is warm and wet. Max Tegmark calculated in 2000 that quantum superpositions in neurons or microtubules would decohere in roughly 10⁻¹³ to 10⁻²⁰ seconds, while neural processes take about 10⁻³ seconds. Supporters dispute his assumptions.

### Are there any quantum effects in living things?

Yes, in specific cases. Enzymes use quantum tunnelling, and the leading explanation of how birds sense Earth's magnetic field involves quantum spin states in cryptochrome proteins. None of this shows that consciousness is quantum.

### Has Penrose's collapse idea been tested?

Partly. A 2020 experiment in the Gran Sasso underground laboratory looked for radiation predicted by the simplest version of gravity-related collapse and did not find it, ruling that version out. Other versions remain open.


---

# Integrated information theory (IIT) explained: Φ, predictions and criticism

Source: https://neuralcosmology.com/en/answers/integrated-information-theory

**What is integrated information theory?** Integrated information theory (IIT), proposed by Giulio Tononi in 2004, says that consciousness is integrated information: how much a system as a whole constrains its own past and future beyond what its parts do, measured by a quantity called Φ. It predicts that consciousness depends on the causal structure of the posterior cortex. It is influential, hard to compute and heavily debated.

## The core idea

Most theories of consciousness start from the brain and ask which activity is conscious. Giulio Tononi started from experience itself. Every experience, he argued, is **integrated** (you see a red ball, not a separate red and a separate round) and **informative** (it is one specific experience out of a vast number possible). A physical system can support consciousness only if it has the same properties.

From this, IIT (*BMC Neuroscience*, 2004; latest version IIT 4.0, Albantakis et al., *PLOS Computational Biology*, 2023) defines consciousness as integrated information and measures it with **Φ**.

## What Φ measures

Take a system of interacting parts. Ask how much its current state constrains its past and future states as a whole, and compare that with what you would get if you cut it into independent pieces along the weakest seam. The difference is Φ.

- A pile of sand has almost no Φ: cutting it changes nothing.
- A feed-forward circuit, where signals flow one way, has zero Φ however complex it is.
- A densely recurrent network, where parts constantly influence each other, can have high Φ.

IIT also says which part of a system is conscious: the subset with maximal Φ, called the main complex.

## Predictions

- **Posterior cortex.** IIT predicts that the core of experience lies in densely interconnected posterior regions of the cortex, a "posterior hot zone", more than in the frontal areas that report and plan.
- **Loss of consciousness.** In deep sleep and anaesthesia, integration should break down even if activity continues. The perturbational complexity index (Casali et al., *Science Translational Medicine*, 2013) tests this: a magnetic pulse to the cortex produces a complex, widespread echo in waking and dreaming, and a simple local one in dreamless sleep and anaesthesia.
- **Machines.** Because Φ depends on physical causal structure, a conventional computer running any program has very low Φ.

## The tests against global workspace theory

A consortium of labs (Cogitate) designed preregistered experiments in which IIT and global neuronal workspace theory predict different outcomes. The results, published in *Nature* in 2025, supported some predictions of each and contradicted others: the sustained synchronisation within posterior cortex that IIT predicted did not appear. The exercise changed the field more than the verdict did: theories of consciousness now compete on predictions that can fail.

## Criticism

- **Computability.** Exact Φ requires checking every way of cutting a system, which is impossible for a brain.
- **Counterintuitive cases.** Scott Aaronson showed in 2014 that simple grid-like arrangements of logic gates can have Φ far higher than a human brain, which many take as a reductio.
- **Testability.** In 2023 more than a hundred researchers called IIT pseudoscience in an open letter; others replied that it makes more testable predictions than most rivals. The disagreement is still live.

## Alternatives to Φ

Several groups propose other integration measures that are cheaper to compute. The *Pointer Architecture* preprint on this site defines a substrate-side measure, **Φ_PA**, which is non-zero only for an observer that reads its own state back, with non-trivial scope and lifespan. It predicts, for example, zero for a single transformer forward pass and a positive value when the model reuses its own cache. The predictions are so far checked on toy encodings. Details are on the [preprint page](https://neuralcosmology.com/en/science/pointer-architecture).

## Bottom line

IIT is the most mathematically developed theory of consciousness. It explains why integration matters and has produced a useful clinical tool. Its central quantity cannot be computed for real brains, and some of its implications strike many researchers as absurd. It is a serious theory that is still under active test.

## FAQ

### What does Φ (phi) measure?

Φ measures how much a system's cause–effect structure is irreducible: how much would be lost if you cut the system into independent parts. A system with Φ = 0 is just a collection of parts; a system with high Φ acts as an integrated whole.

### Is IIT used in medicine?

An offshoot is. The perturbational complexity index (PCI), developed by Marcello Massimini's group, stimulates the cortex with a magnetic pulse and measures how complex the EEG response is. It separates wakefulness, sleep, anaesthesia and disorders of consciousness and is being used to assess unresponsive patients.

### Why is IIT controversial?

Φ cannot be computed for real brains, some simple lattice-like systems would have enormous Φ, and the theory implies consciousness in many systems people find implausible. In 2023 over a hundred researchers signed a letter calling it pseudoscience, which many others rejected.

### Can AI be conscious according to IIT?

IIT ties consciousness to physical causal structure, not to behaviour or software. On standard digital hardware, IIT predicts very low Φ regardless of how intelligent the program seems.


---

# What is dark matter? Evidence, candidates and the search so far

Source: https://neuralcosmology.com/en/answers/what-is-dark-matter

**What is dark matter?** Dark matter is the name for mass we detect only through its gravity: it holds galaxies and clusters together, bends light and shaped the early universe, yet emits no light. It makes up about 85% of all matter. What it is made of is unknown; no particle has been detected despite decades of searches.

## Mass we can weigh but cannot see

Astronomers weigh things by how they move. A planet's speed around the Sun tells you the Sun's mass; a star's speed around a galaxy tells you how much mass lies inside its orbit. When this is done for galaxies and clusters, the answer comes out far larger than all the stars, gas and dust put together. The missing part is called dark matter: it pulls with gravity but neither emits nor absorbs light.

## How we know it is there

The evidence comes from several independent directions, and that is why most physicists take it seriously.

- **Galaxy clusters.** In 1933 Fritz Zwicky measured the speeds of galaxies in the Coma cluster and found them moving so fast that the cluster should fly apart unless it held hundreds of times more mass than its light suggested.
- **Rotation curves.** In the 1970s Vera Rubin and Kent Ford showed that stars at the edges of spiral galaxies orbit as fast as those closer in. Visible matter alone predicts they should slow down. Our answer on [galaxy rotation curves](https://neuralcosmology.com/en/answers/galaxy-rotation-curves-without-dark-matter) covers this in detail.
- **Gravitational lensing.** Mass bends light. Mapping how background galaxies are distorted shows where the mass is, and around galaxies and clusters there is much more of it than we can see.
- **The Bullet Cluster.** In this pair of colliding clusters, the hot gas (most of the ordinary matter) was slowed by the collision, while the mass measured by lensing passed straight through. Douglas Clowe and colleagues published the map in 2006. The mass and the visible matter are in different places.
- **The cosmic microwave background.** The pattern of hot and cold spots left from the early universe depends on how much matter interacts with light and how much does not. The Planck satellite's 2018 fit gives about 5% ordinary matter, 27% dark matter and 68% dark energy.

## What it could be

Nobody knows. The main candidates:

- **WIMPs**, weakly interacting massive particles, which fit naturally into extensions of the Standard Model. For decades they were the favourite.
- **Axions**, very light particles first proposed to solve a different puzzle in the physics of the strong force.
- **Sterile neutrinos**, heavier relatives of ordinary neutrinos that feel only gravity.
- **Primordial black holes**, formed in the first moments after the Big Bang. Observations rule out most mass ranges for them to be all of dark matter.

## The search so far

Detectors deep underground, such as LZ in South Dakota and XENONnT in Italy, wait for a dark matter particle to knock into an atomic nucleus in tonnes of liquid xenon. ADMX listens for axions turning into microwave photons in a strong magnetic field. The Large Hadron Collider looks for energy missing after collisions. None has found anything yet, but each null result narrows where dark matter can hide.

Meanwhile telescopes map dark matter through its gravity: the Euclid space telescope, launched in 2023, and the Vera C. Rubin Observatory measure weak lensing across billions of galaxies.

## Doubts and alternatives

The lack of a detected particle keeps alternatives alive. MOND, proposed by Mordehai Milgrom in 1983, changes the law of gravity at very low accelerations and predicts many galaxy rotation curves from visible matter alone. Erik Verlinde's emergent gravity tries to derive a similar effect from information theory. Both struggle with galaxy clusters and the cosmic microwave background, where dark matter works well. The [rotation curves answer](https://neuralcosmology.com/en/answers/galaxy-rotation-curves-without-dark-matter) compares them.

## Where this programme stands

The research programme on this site does not propose a model of dark matter. Its preprint, *Pointer Architecture* v9.0, works at the level of a formal computational substrate; its one cosmological claim reproduces the standard holographic estimate of dark energy, which concerns the expansion of the universe and says nothing about what dark matter is made of.

## Bottom line

Dark matter is a well-measured effect with an unknown cause. Its gravitational fingerprints agree across galaxies, clusters, lensing and the early universe, which is why it remains the standard explanation. What it is made of is one of the largest open questions in physics.

## FAQ

### What is dark matter in simple terms?

Matter that does not shine or absorb light but still pulls with gravity. We see its effect on stars, galaxies and light, but not the matter itself.

### How much of the universe is dark matter?

According to the Planck satellite's 2018 results, about 27% of the universe's energy content is dark matter, about 5% is ordinary matter and about 68% is dark energy. Dark matter is roughly 85% of all matter.

### What is the difference between dark matter and dark energy?

Dark matter pulls things together through gravity and helps galaxies form. Dark energy is the name for whatever makes the expansion of the universe speed up. They are separate problems with separate evidence.

### Has dark matter been detected directly?

No. Underground detectors such as LZ and XENONnT, axion searches such as ADMX and collider experiments at the LHC have found no dark matter particle so far. They have ruled out large ranges of possible masses and interaction strengths.

### Who discovered dark matter?

Fritz Zwicky inferred unseen mass in the Coma galaxy cluster in 1933. Vera Rubin and Kent Ford's measurements of galaxy rotation in the 1970s turned the idea into a mainstream problem.


---

# The Celestial Code

Source: https://neuralcosmology.com/en/books/celestial-code

*The network of galaxies and the network of neurons in the brain turned out to have similar connection statistics. The book works through this and other findings from peer-reviewed papers and puts forward a hypothesis together with the results that would refute it.*

The first non-fiction book of the series. The author, an AI engineer, brings together work on the large-scale structure of the universe, cellular bioelectricity (Michael Levin), the universe as a learning network (Vitaly Vanchurin), integrated information theory (Giulio Tononi) and the interface theory of perception (Donald Hoffman). Every claim is tied to a published paper, and his own hypothesis, Pointer Architecture, comes with a list of results that would prove it wrong. Between the chapters run autobiographical interludes: a cadet school, a Masonic lodge, the BOOM and AfrikaBurn festivals.

211 pages · every chapter online

Online access to this book: $15 / 690 ₽ / R$ 49

## Chapters

- [From the Author](https://neuralcosmology.com/en/read/celestial-code/preface.md) (free, 2 min)
- [1. A Colander for Soup](https://neuralcosmology.com/en/read/celestial-code/ch01.md) (free, 16 min)
- [Strangled at Birth](https://neuralcosmology.com/en/read/celestial-code/interlude1.md) (free, 13 min)
- [2. Twenty-Seven Zeros](https://neuralcosmology.com/en/read/celestial-code/ch02.md) (free, 14 min)
- [A Boy from Stavropol](https://neuralcosmology.com/en/read/celestial-code/interlude5.md) (free, 19 min)
- [3. A Photo Costs Energy](https://neuralcosmology.com/en/read/celestial-code/ch03.md) (free, 17 min)
- [Cosmos](https://neuralcosmology.com/en/read/celestial-code/interlude6.md) (free, 11 min)
- [4. Four Rules](https://neuralcosmology.com/en/read/celestial-code/ch_comp.md) (opening open, 16 min)
- [Pointers](https://neuralcosmology.com/en/read/celestial-code/interlude8.md) (opening open, 5 min)
- [5. The Universe Learns](https://neuralcosmology.com/en/read/celestial-code/ch04.md) (opening open, 18 min)
- [Remix](https://neuralcosmology.com/en/read/celestial-code/interlude2.md) (opening open, 11 min)
- [6. A Body Without a Manual](https://neuralcosmology.com/en/read/celestial-code/ch05.md) (opening open, 17 min)
- [The Day Before](https://neuralcosmology.com/en/read/celestial-code/interlude3.md) (opening open, 5 min)
- [7. The Brain Lies](https://neuralcosmology.com/en/read/celestial-code/ch06.md) (opening open, 19 min)
- [Three Wives](https://neuralcosmology.com/en/read/celestial-code/interlude10.md) (opening open, 14 min)
- [8. Ninety-Nine Percent](https://neuralcosmology.com/en/read/celestial-code/ch07.md) (opening open, 17 min)
- [Burns](https://neuralcosmology.com/en/read/celestial-code/interlude11.md) (opening open, 11 min)
- [9. Sixty Times](https://neuralcosmology.com/en/read/celestial-code/ch08.md) (opening open, 17 min)
- [Coincidence](https://neuralcosmology.com/en/read/celestial-code/interlude9.md) (opening open, 10 min)
- [10. Seven Knives in the Back](https://neuralcosmology.com/en/read/celestial-code/ch09.md) (opening open, 13 min)
- [Cape Town](https://neuralcosmology.com/en/read/celestial-code/interlude12.md) (opening open, 8 min)
- [11. The Archive Is Heavier Than Stars](https://neuralcosmology.com/en/read/celestial-code/ch11.md) (opening open, 67 min)
- [The Architecture of Mystery](https://neuralcosmology.com/en/read/celestial-code/interlude7.md) (opening open, 9 min)
- [12. Where Wisdom Is Buried](https://neuralcosmology.com/en/read/celestial-code/ch_wisdom.md) (opening open, 19 min)
- [Datsan](https://neuralcosmology.com/en/read/celestial-code/interlude13.md) (opening open, 8 min)
- [13. Beneath the Floor — Another Floor](https://neuralcosmology.com/en/read/celestial-code/ch10.md) (opening open, 13 min)
- [Reassembly](https://neuralcosmology.com/en/read/celestial-code/interlude4.md) (opening open, 13 min)
- [Afterword](https://neuralcosmology.com/en/read/celestial-code/afterword.md) (opening open, 8 min)
- [Postscript](https://neuralcosmology.com/en/read/celestial-code/postscript.md) (paid, 4 min)
- [About the Author](https://neuralcosmology.com/en/read/celestial-code/about.md) (paid, 1 min)


---

# Conscious Selection

Source: https://neuralcosmology.com/en/books/conscious-selection

*The eye keeps only a spot the size of a fingernail in sharp focus, and a memory is rewritten every time it is recalled. The second book counts how many times the world is filtered before you choose anything, and reports what was left of the author's hypothesis after testing.*

The second non-fiction book of the series. Yarbus showed a subject the same Repin painting seven times and recorded seven different paths of the gaze, changing only the question asked beforehand. Nader showed that a recalled memory has to be consolidated again. Zurek explained why the environment leaves a quantum system only a few stable states. The book lines these results up into one argument: selection runs many times before a person gets to choose. At the end the author puts his own hypothesis from The Celestial Code through the same test and tells three stories of things that did not work, including a calculation of his that missed the Planck data by a hundred sigma.

Draft in proofreading · every chapter online

Online access to this book: $15 / 690 ₽ / R$ 49

## Chapters

- [From the Author](https://neuralcosmology.com/en/read/conscious-selection/preface.md) (free, 4 min)
- [1. They Did Not Expect Him](https://neuralcosmology.com/en/read/conscious-selection/ch01.md) (free, 11 min)
- [2. Twenty Watts](https://neuralcosmology.com/en/read/conscious-selection/ch02.md) (free, 8 min)
- [3. The Ceiling](https://neuralcosmology.com/en/read/conscious-selection/ch03.md) (opening open, 8 min)
- [4. From Scratch Every Time](https://neuralcosmology.com/en/read/conscious-selection/ch04.md) (opening open, 8 min)
- [The Round Lamp](https://neuralcosmology.com/en/read/conscious-selection/interlude4.md) (opening open, 6 min)
- [5. So As Not to Be Surprised](https://neuralcosmology.com/en/read/conscious-selection/ch05.md) (opening open, 10 min)
- [The Valley of Flowers](https://neuralcosmology.com/en/read/conscious-selection/interlude5.md) (paid, 3 min)
- [6. What Outlives the Gaze](https://neuralcosmology.com/en/read/conscious-selection/ch06.md) (opening open, 7 min)
- [The Spider](https://neuralcosmology.com/en/read/conscious-selection/interlude6.md) (opening open, 6 min)
- [7. How Many Steps](https://neuralcosmology.com/en/read/conscious-selection/ch07.md) (opening open, 7 min)
- [The Trading House](https://neuralcosmology.com/en/read/conscious-selection/interlude7.md) (paid, 4 min)
- [8. Three Conditions](https://neuralcosmology.com/en/read/conscious-selection/ch08.md) (opening open, 7 min)
- [The Truck's Steering Wheel](https://neuralcosmology.com/en/read/conscious-selection/interlude8.md) (opening open, 8 min)
- [9. Forty Years and It Won't Start](https://neuralcosmology.com/en/read/conscious-selection/ch09.md) (opening open, 7 min)
- [Tokens](https://neuralcosmology.com/en/read/conscious-selection/interlude9.md) (paid, 4 min)
- [10. With My Own Hands](https://neuralcosmology.com/en/read/conscious-selection/ch10.md) (opening open, 8 min)
- [Two Pairs of Trousers](https://neuralcosmology.com/en/read/conscious-selection/interlude10.md) (paid, 5 min)
- [11. Fifty-Two Cycles](https://neuralcosmology.com/en/read/conscious-selection/ch11.md) (opening open, 7 min)
- [The Record Book](https://neuralcosmology.com/en/read/conscious-selection/interlude11.md) (opening open, 6 min)
- [12. Graph Paper](https://neuralcosmology.com/en/read/conscious-selection/ch12.md) (opening open, 9 min)
- [Dubai](https://neuralcosmology.com/en/read/conscious-selection/interlude12.md) (opening open, 5 min)
- [13. Who Asks the Question](https://neuralcosmology.com/en/read/conscious-selection/ch13.md) (opening open, 7 min)
- [Clever Thoughts](https://neuralcosmology.com/en/read/conscious-selection/interlude13.md) (paid, 1 min)
- [Afterword](https://neuralcosmology.com/en/read/conscious-selection/afterword.md) (opening open, 8 min)


---

# Bugs Academy

Source: https://neuralcosmology.com/en/books/bugs-academy

*Moscow physicist Dan Reinov sees a car crash take shape a second before it happens. Then a man who is seven hundred years old calls and explains what really happened to his mother.*

Daniil Reinov is twenty-eight, a quantum physicist in Moscow, and since childhood he has seen events assemble a second before they happen. Doctors called it apophenia. When the qubits in his lab start responding to his thoughts, he is found by Wei Lin, a teacher who is seven hundred years old, and taken to an Academy that looks like a country dacha outside Moscow. Four students like him study there, reality turns out to have six layers, and Dan's mother, who vanished twenty-one years ago, is alive in another body. A novel about immortality and what it costs.

Finished novel · 232 pages · seeking a publisher

Online access to this book: $15 / 690 ₽ / R$ 49

## Chapters

- [Preface](https://neuralcosmology.com/en/read/bugs-academy/preface.md) (free, 1 min)
- [1. Peripheral Vision](https://neuralcosmology.com/en/read/bugs-academy/ch01.md) (free, 23 min)
- [2. The Thing That Doesn't Happen Fourteen Times in a Row](https://neuralcosmology.com/en/read/bugs-academy/ch02.md) (free, 23 min)
- [3. A Debt](https://neuralcosmology.com/en/read/bugs-academy/ch03.md) (free, 20 min)
- [4. An Inhuman Gaze](https://neuralcosmology.com/en/read/bugs-academy/ch04.md) (free, 18 min)
- [5. How to Properly Lose Your Mind](https://neuralcosmology.com/en/read/bugs-academy/ch05.md) (free, 25 min)
- [6. People Who Don't Smile](https://neuralcosmology.com/en/read/bugs-academy/ch06.md) (opening open, 15 min)
- [7. Six Floors Down](https://neuralcosmology.com/en/read/bugs-academy/ch07.md) (opening open, 15 min)
- [8. Why Everything Looks Like Everything](https://neuralcosmology.com/en/read/bugs-academy/ch08.md) (opening open, 16 min)
- [9. Silent Weapons](https://neuralcosmology.com/en/read/bugs-academy/ch09.md) (opening open, 17 min)
- [10. Who Lives Inside Your Head](https://neuralcosmology.com/en/read/bugs-academy/ch10.md) (opening open, 25 min)
- [11. On the Usefulness of Wings](https://neuralcosmology.com/en/read/bugs-academy/ch11.md) (opening open, 24 min)
- [12. Four Ways to Be Wrong](https://neuralcosmology.com/en/read/bugs-academy/ch12.md) (opening open, 16 min)
- [Pancakes](https://neuralcosmology.com/en/read/bugs-academy/interlude.md) (opening open, 12 min)
- [13. Lab Rats](https://neuralcosmology.com/en/read/bugs-academy/ch13.md) (opening open, 17 min)
- [14. Immortals Who Don't Care](https://neuralcosmology.com/en/read/bugs-academy/ch14.md) (opening open, 15 min)
- [15. God as a Job Description](https://neuralcosmology.com/en/read/bugs-academy/ch15.md) (opening open, 17 min)
- [16. Who Herds the Shepherds](https://neuralcosmology.com/en/read/bugs-academy/ch16.md) (opening open, 21 min)
- [17. The Incubator](https://neuralcosmology.com/en/read/bugs-academy/ch17.md) (opening open, 18 min)
- [18. You Are — Who?](https://neuralcosmology.com/en/read/bugs-academy/ch18.md) (opening open, 21 min)
- [19. Cocoon](https://neuralcosmology.com/en/read/bugs-academy/ch19.md) (opening open, 15 min)
- [20. Those Who Didn't Fully Die](https://neuralcosmology.com/en/read/bugs-academy/ch20.md) (opening open, 14 min)
- [21. Mother's Handwriting](https://neuralcosmology.com/en/read/bugs-academy/ch21.md) (opening open, 16 min)
- [22. What Does a Caterpillar Know About Flight](https://neuralcosmology.com/en/read/bugs-academy/ch22.md) (opening open, 20 min)
- [23. Three Recipes for Human Happiness](https://neuralcosmology.com/en/read/bugs-academy/ch23.md) (opening open, 18 min)
- [24. The One You Were Before Yourself](https://neuralcosmology.com/en/read/bugs-academy/ch24.md) (opening open, 20 min)
- [25. God's Rough Draft](https://neuralcosmology.com/en/read/bugs-academy/ch25.md) (opening open, 13 min)
- [Glossary of Academy Terms](https://neuralcosmology.com/en/read/bugs-academy/glossary.md) (paid, 2 min)


---

# Era of Architects

Source: https://neuralcosmology.com/en/books/era-of-architects

*Moscow, 2033. The state opens an Institute of Awakening and licenses expanded perception, while Dan Reinov's seven-year-old daughter has had that perception since birth.*

The sequel to Bugs Academy, seven years later. In the Moscow of 2033 the state certifies people who see reality as code; the system is run by Dan's former friend Artyom, and his old enemy Iero blows it up on prime-time television. Dan's daughter Lia was born with expanded perception, and the seven-hundred-year-old teacher prepares her for a decision she will have to make for the whole species fifteen years later. Behind them stand the archives of six civilisations that never got this far. The story is closed by an epilogue.

Finished novel · 246 pages

Online access to this book: $15 / 690 ₽ / R$ 49

## Chapters

- [Preface](https://neuralcosmology.com/en/read/era-of-architects/preface.md) (free, 1 min)
- [1. The Porch](https://neuralcosmology.com/en/read/era-of-architects/ch01.md) (free, 31 min)
- [2. The First Wave](https://neuralcosmology.com/en/read/era-of-architects/ch02.md) (free, 30 min)
- [3. The Thread Is Calling](https://neuralcosmology.com/en/read/era-of-architects/ch03.md) (opening open, 31 min)
- [Protocol P-7/DV](https://neuralcosmology.com/en/read/era-of-architects/interlude1.md) (opening open, 6 min)
- [4. Elsa](https://neuralcosmology.com/en/read/era-of-architects/ch04.md) (opening open, 26 min)
- [5. Karl](https://neuralcosmology.com/en/read/era-of-architects/ch05.md) (opening open, 23 min)
- [6. Macro-contact](https://neuralcosmology.com/en/read/era-of-architects/ch06.md) (opening open, 19 min)
- [7. Chronicles](https://neuralcosmology.com/en/read/era-of-architects/ch07.md) (opening open, 21 min)
- [Lia's Essay: Winter](https://neuralcosmology.com/en/read/era-of-architects/interlude2.md) (opening open, 6 min)
- [8. Iero Returns](https://neuralcosmology.com/en/read/era-of-architects/ch08.md) (opening open, 22 min)
- [9. The Ordo Catalog](https://neuralcosmology.com/en/read/era-of-architects/ch09.md) (opening open, 22 min)
- [10. Alpha Partners](https://neuralcosmology.com/en/read/era-of-architects/ch10.md) (opening open, 29 min)
- [11. The Heptarch's Visit](https://neuralcosmology.com/en/read/era-of-architects/ch11.md) (opening open, 32 min)
- [The Markova Declaration](https://neuralcosmology.com/en/read/era-of-architects/interlude3.md) (opening open, 12 min)
- [12. Failures](https://neuralcosmology.com/en/read/era-of-architects/ch12.md) (opening open, 27 min)
- [13. Timewalk](https://neuralcosmology.com/en/read/era-of-architects/ch13.md) (opening open, 25 min)
- [14. A Response Is Needed](https://neuralcosmology.com/en/read/era-of-architects/ch14.md) (opening open, 17 min)
- [Irene's Voice](https://neuralcosmology.com/en/read/era-of-architects/interlude4.md) (opening open, 8 min)
- [15. Three Answers](https://neuralcosmology.com/en/read/era-of-architects/ch15.md) (opening open, 20 min)
- [16. Heptarch](https://neuralcosmology.com/en/read/era-of-architects/ch16.md) (opening open, 15 min)
- [17. The Era](https://neuralcosmology.com/en/read/era-of-architects/ch17.md) (opening open, 13 min)
- [Fragment of the Book of Life](https://neuralcosmology.com/en/read/era-of-architects/interlude5.md) (paid, 1 min)
- [Epilogue](https://neuralcosmology.com/en/read/era-of-architects/epilog.md) (opening open, 12 min)


---

# Slime mold Physarum builds a network: simulation

Source: https://neuralcosmology.com/en/experiments/cc-ch02-physarum

Jeff Jones’s model (2010): nine thousand particles sense the trail ahead, turn toward the strongest and leave their own. The algorithm Burchett and Elek used to recover the filaments of the cosmic web works the same way.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/cc-ch02-physarum
From the book *The Celestial Code*, chapter [Twenty-Seven Zeros](https://neuralcosmology.com/en/read/celestial-code/ch02.md).

> The argument here is about something else: the rules of optimization are the same everywhere, at every scale, from slime mold to supercluster — simply because no other optimum exists in principle, and the mathematics of that optimization is one and the same, in no way dependent on the substrate.
>
> If in Japan — where Physarum polycephalum, by the way, was used to optimize the Tokyo rail network (Tero et al., Science, 2010) — you told the scientists, “Your slime mold is designing the cosmos” — they’d probably laugh. And they’d be wrong to.

## What the slime mold Physarum is

Physarum polycephalum is a yellow slime mold that lives on rotting leaves and bark in damp woods. Its active form, the plasmodium, is one giant cell with millions of nuclei that can spread over tens of centimetres. It has no neurons and no brain, yet it finds shortest paths.

In 2000 Toshiyuki Nakagaki put pieces of the slime mold into a plastic maze with oat flakes at the two exits. The mold first filled the whole maze, then withdrew from the dead ends and left a single thick tube along the shortest route between the food. The paper appeared in Nature.

In 2010 Atsushi Tero, Nakagaki and colleagues laid out oat flakes like the cities around Tokyo and placed the mold at the centre. Within about a day it linked them with a network comparable to the real railway in length, resilience and cost (Science, 2010). The trick is simple: tubes that carry more flow thicken, and the rest wither.

In 2010 Jeff Jones captured this behaviour with a model of thousands of particles: each senses the trail ahead, turns toward the strongest and leaves its own. That model is what this experiment runs. In 2020 Joseph Burchett, Oskar Elek and colleagues applied an algorithm based on it to 37,000 galaxies and recovered the filaments of the cosmic web between them, including where they cannot be seen directly.

## FAQ

### Does a slime mold have a brain?

No. The Physarum plasmodium is a single cell with many nuclei and no neurons. It finds paths because tubes with strong flow thicken while weak ones wither.

### Did a slime mold really reproduce the Tokyo railway?

Yes. In Tero and Nakagaki's experiment (Science, 2010) the mold linked oat flakes laid out like the cities around Tokyo with a network comparable to the real one in length, fault tolerance and cost.

### What does it have to do with the cosmic web?

In 2020 astronomers applied an algorithm based on the slime mold model to 37,000 SDSS galaxies and recovered the gas filaments between them, then checked the result against absorption in the spectra of distant quasars.
- [What is the cosmic web? Filaments, voids and the largest structure in the universe](https://neuralcosmology.com/en/answers/cosmic-web)
- [Is the universe a neural network? What the evidence says](https://neuralcosmology.com/en/answers/is-the-universe-a-neural-network)


---

# Landauer's principle: the cost of erasing a bit

Source: https://neuralcosmology.com/en/experiments/cc-ch03-landauer

Landauer’s limit: kT ln 2 for every erased bit, where k is Boltzmann’s constant and T the temperature of the surroundings. Bérut and colleagues measured it in 2012 with a single glass bead in a double laser trap.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/cc-ch03-landauer
From the book *The Celestial Code*, chapter [A Photo Costs Energy](https://neuralcosmology.com/en/read/celestial-code/ch03.md).

> And here’s why it matters: if erasing information produces a physical effect, then information is physical directly and literally; it is physics. A bit is just as much a physical quantity as a joule or a meter.
>
> The consequence you want to wave away but can’t: every time you delete your ex’s photo from your phone, the Universe pays an energy cost. A tiny one — 3×10−21 joules per bit — but a real one. Information doesn’t disappear for free — not anywhere, not ever, not even on your phone.


---

# Double-slit experiment: interactive simulation

Source: https://neuralcosmology.com/en/experiments/cc-ch03-double-slit

Each photon leaves a single dot on the screen. Without a detector the dots follow the square of the sum of two amplitudes, with fringes; with a detector at the slits, the sum of two probabilities, with none. The setup is simplified; the probabilities are the textbook ones.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/cc-ch03-double-slit
From the book *The Celestial Code*, chapter [A Photo Costs Energy](https://neuralcosmology.com/en/read/celestial-code/ch03.md).

> Non-locality of the information graph — that’s the whole explanation.
>
> Wheeler called this the “participatory universe”: the observer doesn’t passively record reality but participates in creating it. Every act of observation is an answer to a yes-or-no question, and that answer is a bit. And from these bits, reality is assembled.
>
> It from bit.

## What the double-slit experiment shows

Light passes through two narrow parallel slits and lands on a screen. If light were a stream of particles, you would see two bright bands opposite the slits. Instead you see many bands, bright and dark in turn. That is how waves add up: where crests meet crests the light grows stronger, where a crest meets a trough it cancels. Thomas Young showed this in the early 1800s, and the experiment became the main evidence that light behaves as a wave.

The strange part begins when particles go through one at a time. In 1989 Akira Tonomura's group at Hitachi sent electrons through the apparatus singly. Each electron left a single dot on the screen, like a particle. After a few thousand dots, though, the same fringes appeared, as if every electron had passed through both slits and interfered with itself.

Put a detector at the slits that records which one a particle went through, and the fringes disappear, leaving two blurred bands. Quantum mechanics describes this without mysticism: while the path is unrecorded, the amplitudes of the two paths add and the pattern is there; once the path is recorded anywhere, the probabilities add and the pattern is gone. No human observer has to look. Any physical trace of the path is enough.

In 1999 Markus Arndt and Anton Zeilinger obtained the same fringes with C60 fullerene molecules, sixty carbon atoms each. Heat those molecules and they emit thermal photons that give away their path, and the fringes fade. The experiment on hot fullerenes shows it.

## FAQ

### Why do the fringes disappear when a detector is added?

The detector records which slit the particle went through. Once the path is known, the amplitudes of the two paths no longer add; the probabilities do, and a sum of probabilities has no fringes. No person needs to look: it is enough that information about the path is left in the device or the surroundings.

### Has the experiment been done with single particles?

Yes. Claus Jönsson did it with electrons in 1961, and in 1989 Akira Tonomura's group sent electrons strictly one at a time and filmed the fringes building up from individual dots.

### Does it work with large molecules?

Yes. In 1999 Anton Zeilinger's group observed interference of C60 fullerene molecules, and later experiments used molecules of thousands of atoms.
- [Is consciousness quantum? Orch-OR, decoherence and what experiments show](https://neuralcosmology.com/en/answers/is-consciousness-quantum)
- [Landauer's principle: why erasing information costs energy](https://neuralcosmology.com/en/answers/landauer-principle)


---

# Starling murmuration from three rules: simulation

Source: https://neuralcosmology.com/en/experiments/cc-ch04-murmuration

Craig Reynolds’s model (1987): three rules and no leader. Real starlings, Ballerini and colleagues showed in 2008, track about six or seven nearest neighbours, however many birds are close by.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/cc-ch04-murmuration
From the book *The Celestial Code*, chapter [The Universe Learns](https://neuralcosmology.com/en/read/celestial-code/ch04.md).

> Tononi would say a different one (GPT has a low Φ due to its modular architecture). Vanchurin — that it depends on the scale. Hoffman — that we cannot know, because we see only the interface.
>
> I would answer differently: the question is set up wrong, and the correct one runs like this — at what depth in the stack? Perhaps GPT and the brain are both emergence, but at different levels of the same graph: one at the silicon level, the other at the protein level, and both are nodes in one network.


---

# Two-headed planaria: Michael Levin's experiment

Source: https://neuralcosmology.com/en/experiments/cc-ch05-planaria

A diagram of the experiment, not a simulation. Colour shows the bioelectric pattern: which end becomes a head. Every worm has the same genes; only the pattern changes.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/cc-ch05-planaria
From the book *The Celestial Code*, chapter [A Body Without a Manual](https://neuralcosmology.com/en/read/celestial-code/ch05.md).

> (You’re thinking right now: “That’s just physics. The shape is determined by hydrodynamics, surface tension, mechanics.” Maybe. But then explain to me why this particular physics leads to that shape capable of reproducing — to a Pac-Man with an opening of the right size, rather than a circle or a cube. Physics explains how, but says nothing about what for.)
>
> Nobody told the cells — mathematics did. The same principle as in the slime mold building an optimal network; the same self-organization that at other scales builds the cosmic web. Form is computed from rules, from physics, from what Kauffman called “order for free,” and is not assigned by anybody.
>
> How did they “know” what to build?

## What planaria are and how they regenerate

Planarians are freshwater flatworms from a few millimetres to a couple of centimetres long, in the phylum Platyhelminthes. They have two light-sensitive eyespots on the head that give them a permanently surprised look. They live under stones in streams and ponds and feed on small invertebrates.

Their signature trait is regeneration. Cut a planarian into pieces and each piece grows into a whole worm with a head, a brain and a tail. As early as 1898 Thomas Hunt Morgan showed that a fragment of 1/279 of the body is enough. The work is done by neoblasts, stem cells spread throughout the body: in 2011 Peter Reddien's lab showed that a single such cell can rebuild an entire worm.

How does a fragment know where its head should go? Michael Levin's lab at Tufts University showed that the information is carried by a bioelectric pattern, the voltage differences across cell membranes along the body. Briefly change that pattern with a drug that blocks the gap junctions between cells, and the fragment grows two heads, one at each end.

The most surprising part came next. The two-headed worms were cut again, with no drug at all, and the pieces again grew into two-headed worms. Their genes were unchanged. What changed was a record in the electrical pattern, and the body kept reading it. The diagram on this page shows exactly that: colour marks the pattern, and the worm's shape follows it.

## FAQ

### What is a planarian?

A freshwater flatworm up to a couple of centimetres long, in the phylum Platyhelminthes, known for regrowing its entire body from a small fragment.

### How do planarians regenerate?

Through neoblasts, stem cells scattered throughout the body. They divide and turn into the missing tissues, while bioelectric and chemical signals tell them what to build where.

### How were two-headed planarians made?

In Michael Levin's lab, worm fragments were briefly treated with a substance that blocks gap junctions, which changed the electrical pattern. Those fragments grew two heads, and the worms stayed two-headed through later cuts with no change to their genes.
- [Michael Levin and bioelectricity: how cells store the plan of the body](https://neuralcosmology.com/en/answers/michael-levin-bioelectricity)


---

# Sixth in your pocket

Source: https://neuralcosmology.com/en/experiments/cc-ch11-sixth

A pocket version: the stack words and the substrate words from LANGUAGE.md, with nodes numbered from one as in the real Sixth. The rung programs come from examples in github.com/neuralcosmology/sixth, which also holds the full Racket implementation.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/cc-ch11-sixth
From the book *The Celestial Code*, chapter [The Archive Is Heavier Than Stars](https://neuralcosmology.com/en/read/celestial-code/ch11.md).

> State evolution alone isn’t enough here, — you need law evolution. Without it you get an automaton: inside the initial grammar it can do a great deal, but it cannot in principle step beyond its limits. With law evolution you get a recursive reality: the grammar grows by itself, and floor by floor the system acquires new kinds of what is able to happen at all. State evolution changes what is right now; law evolution changes what is able to happen next at all; a self-rewriting substrate couples both processes into one device.
>
> Without this move, Pointer Architecture is easy to confuse on first reading with yet another simulation, — and the whole construction then comes down to a decorative calculator with four buttons and pretty marketing brochures around it.


---

# Conway's Game of Life online

Source: https://neuralcosmology.com/en/experiments/cc-comp-life

Gosper’s glider gun (1970) was the first pattern found to grow forever: every thirty generations it fires another glider. Conway had offered fifty dollars for such a pattern, and paid.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/cc-comp-life
From the book *The Celestial Code*, chapter [Four Rules](https://neuralcosmology.com/en/read/celestial-code/ch_comp.md).

> Conway, for the record, hated his “Game of Life.” His whole life. It annoyed him that his other work — serious mathematics he’d devoted decades to — drowned in the shadow of a cellular automaton he’d come up with over tea. As if Pushkin had become famous not for Eugene Onegin but for a doodle in the margins of a notebook. But that “doodle” turned out to be a door. Through it, mathematics saw what was previously thought impossible: from simple rules — without design, without instructions, without anyone’s “mind” — complexity, structure, motion emerge. An entire world.
>
> After which a question hung in the air that hasn’t been put to bed since 1970: what if our world is a cellular automaton too, just with more complex rules and a finer grid?

## What Conway's Game of Life is

The Game of Life was invented by the English mathematician John Horton Conway in 1970 and made famous by Martin Gardner's column in Scientific American that October. It is a zero-player game: you only place live cells on an infinite grid, and everything after that happens by itself, generation after generation.

There are four rules. A live cell with fewer than two live neighbours dies of loneliness. A live cell with two or three neighbours lives on. A live cell with more than three neighbours dies of overcrowding. An empty cell with exactly three live neighbours comes to life. The neighbours are the eight cells around it.

A surprising amount comes out of these rules. There are still lifes that never change, oscillators that blink in place, and gliders, five cells that crawl diagonally across the grid. Conway offered fifty dollars to anyone who could find a pattern that grows forever. Bill Gosper of MIT found one in 1970: his glider gun fires a new glider every thirty generations.

It was later shown that any computer can be built inside Life: it is Turing complete. That is why it became the favourite example of complex behaviour arising from simple local rules with no plan from above.

## FAQ

### What are the rules of the Game of Life?

A live cell survives if it has two or three live neighbours and dies otherwise. An empty cell becomes alive if it has exactly three live neighbours. Neighbours are the eight surrounding cells, and all cells update at once.

### Who invented the Game of Life?

The English mathematician John Horton Conway, in 1970. Martin Gardner introduced it to a wide audience in Scientific American.

### Can you build a computer in the Game of Life?

Yes. Life is Turing complete: logic gates can be assembled from gliders and guns, and in 2000 Paul Rendell built a Turing machine inside it.
- [Simulation theory: are we living in a simulation, or is the universe computing itself?](https://neuralcosmology.com/en/answers/is-the-universe-a-simulation)
- [Is the universe a neural network? What the evidence says](https://neuralcosmology.com/en/answers/is-the-universe-a-neural-network)


---

# Elementary cellular automata: all 256 rules

Source: https://neuralcosmology.com/en/experiments/cc-comp-rule110

A rule’s number is its eight answers read as a binary number. Rule 110’s universality was proved by Matthew Cook, who worked for Wolfram; the proof was published in 2004.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/cc-comp-rule110
From the book *The Celestial Code*, chapter [Four Rules](https://neuralcosmology.com/en/read/celestial-code/ch_comp.md).

> The scientific community ignored the book. Twelve hundred and eighty pages without peer review, written by a man who hadn’t talked to anyone for ten years, — how do you take that seriously, — even if the content is revolutionary, — especially if it’s revolutionary. Medicine has been through this before: in the nineteenth century, a doctor named Semmelweis was fired and committed to an asylum for suggesting that doctors wash their hands before surgery (more on that in chapter nine). When data is inconvenient, the one to blame turns out to be the messenger.
>
> Wolfram waited — exactly eighteen years, during which the scientific community slowly and without any enthusiasm grew used to the thought that simple rules can produce complexity.


---

# Yarbus eye-tracking experiment: try it

Source: https://neuralcosmology.com/en/experiments/cs-ch01-yarbus

Yarbus’s experiment on Repin’s “They Did Not Expect Him”, with a cursor in place of the suction cap on the eye. Ask yourself a question, look for twenty seconds, then take another and lay the paths over each other. Nothing is sent anywhere.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/cs-ch01-yarbus
From the book *Conscious Selection*, chapter [They Did Not Expect Him](https://neuralcosmology.com/en/read/conscious-selection/ch01.md).

> Three minutes is quite a long time if the only thing required of you is to look at one painting: the eye manages several hundred fixations in that time. And out of these several hundred fixations, which could have been spread over the canvas in any way at all, the subject each time assembled a new combination, tailored to the sentence he had heard a second earlier.
>
> This was published in 1965 in the book The Role of Eye Movements in the Process of Vision, came out in English two years later, and since then Yarbus’s drawings have been reprinted, it seems, by everyone who has ever had to give a lecture on attention.


---

# Blind spot test: find your own

Source: https://neuralcosmology.com/en/experiments/cs-ch02-blind-spot

The spot lies about fifteen degrees from the point of gaze, toward the temple. Edme Mariotte first described it in the 1660s: at that place on the retina the optic nerve leaves the eye.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/cs-ch02-blind-spot
From the book *Conscious Selection*, chapter [Twenty Watts](https://neuralcosmology.com/en/read/conscious-selection/ch02.md).

> This, incidentally, makes clear where the trajectories on Yarbus’s film came from. The eye runs over the canvas for a purely mechanical reason: it sees two degrees sharply at a time, and so, to make out the painting, it has to apply those two degrees to it in several hundred places; the jump of the eye turns out to be a direct consequence of the retina being built unevenly.
>
> Since we are talking about holes in your vision, here is another one. Each eye has a place where the optic nerve exits and there are no light-sensitive cells at all, a blind spot roughly the size of a fist at arm’s length. You have never once seen it in your whole life, because it is neatly plastered over with whatever the system believes must be there.

## What the blind spot in your eye is

Each eye has a place that sees nothing. Where the optic nerve and blood vessels leave the eyeball, the retina has no rods and no cones. This is the optic disc, an oval about five degrees of the visual field wide, roughly fifteen degrees from the point you are looking at, toward the temple.

The French physicist Edme Mariotte described the blind spot in the 1660s. He noticed that if you look with one eye at a certain point, a small object to the side vanishes completely while everything around it stays visible. An anatomical detail became one of the clearest demonstrations of how vision works.

Why don't we see the hole? First, the two eyes cover each other's blind spots. Second, even with one eye closed, the brain fills the gap with whatever surrounds it, continuing lines and stretching backgrounds and textures. What we see is not the raw image from the retina; it is what the brain builds from it.

That is why the blind spot matters when we talk about consciousness. It shows how confidently perception puts a guess where data is missing, without marking the seam. To find yours, close your left eye, look at the cross with your right eye and slowly change your distance from the screen until the dot vanishes. The driver's blind spot, the zone the mirrors don't show, shares only the name.

## FAQ

### Why does the eye have a blind spot?

Where the optic nerve leaves the eye, the retina has no light-sensitive cells. This spot is called the optic disc.

### Why don't we notice the blind spot?

The two eyes cover each other's blind spots, and the brain fills the gap with the surrounding pattern and colour.

### How do I find my blind spot?

Close your left eye, look at the cross with your right eye and slowly move toward or away from the screen. At a certain distance the dot to the side disappears, because its image falls on the optic disc.
- [The hard problem of consciousness: the main theories and what can be tested](https://neuralcosmology.com/en/answers/hard-problem-of-consciousness)


---

# False memory: the Loftus and Palmer experiment

Source: https://neuralcosmology.com/en/experiments/cs-ch04-loftus

Loftus and Palmer, 1974. The original subjects watched real driver-safety films of crashes; this is a diagram. The bars are the paper’s mean speed estimates, converted to km/h.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/cs-ch04-loftus
From the book *Conscious Selection*, chapter [From Scratch Every Time](https://neuralcosmology.com/en/read/conscious-selection/ch04.md).

> Those who got the placebo feared the spiders. So did those who got the drug without a reminder. The only ones who did not were those in whom both coincided, and the fear did not return to them even after they were shocked again. The paper came out in Nature Neuroscience under the title “Beyond extinction.”
>
> There is a detail here that decides everything. The people remembered perfectly well that they had been shocked and could recount the circumstances of the experiment in detail: what vanished was the fear, while the knowledge stayed in place; the drug removed one property from the record without touching the record itself.


---

# Heated molecules lose their fringes

Source: https://neuralcosmology.com/en/experiments/cs-ch06-hot-fullerenes

Hackermüller and colleagues, Nature, 2004. The visibility curve here is a sketch following the chapter: fringes survive up to about 1,500 K and fade out by 3,000 K.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/cs-ch06-hot-fullerenes
From the book *Conscious Selection*, chapter [What Outlives the Gaze](https://neuralcosmology.com/en/read/conscious-selection/ch06.md).

> Think about what exactly extinguished it. The molecule was left completely untouched: nothing was fired at it and no obstacles were put in its way. The one thing that changed was that information about its position became available to the outside. Interference dies at the moment the world gets the possibility of learning the answer, and it makes no difference whatsoever that nobody will ever catch or read these escaped photons.
>
> There is no reader here at all. There is only the availability of the answer, and that is enough.


---

# Assembly theory: calculate an assembly index

Source: https://neuralcosmology.com/en/experiments/cs-ch07-assembly

The chapter’s rules: letters are free, a step joins two pieces already built, and whatever is built stays in stock. For words up to sixteen letters the index is found by search; if the search runs out of budget, an upper bound is shown with “≤”.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/cs-ch07-assembly
From the book *Conscious Selection*, chapter [How Many Steps](https://neuralcosmology.com/en/read/conscious-selection/ch07.md).

> The second quantity in this construction is the number of copies, and it weighs no less than the first. A single complex molecule proves nothing yet: who knows what might have stuck together by chance over a billion years. But a million identical complex molecules means that somewhere a workshop is operating that assembles them to one blueprint.
>
> A high index plus a large print run: that, according to Cronin, is the fingerprint that selection leaves on matter.


---

# Seal a prediction

Source: https://neuralcosmology.com/en/experiments/cs-ch11-envelope

SHA-256, the same fingerprint as in the attestations/ledger.txt log of the Sixth repository. The text and fingerprint stay in your browser.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/cs-ch11-envelope
From the book *Conscious Selection*, chapter [Fifty-Two Cycles](https://neuralcosmology.com/en/read/conscious-selection/ch11.md).

> Over a few years nine rules have grown up around this procedure, of which I will name three. Read the literature before registering the prediction, not after. Do not count fewer than a thousand independent runs as a result. And as a separate item, hard-won from my own mistakes, there is a check for tautology: whether what I am going to measure might turn out to be a simple consequence of how I defined it.
>
> As of today there are fifty-two predictions in the register.


---

# Fourteen times in a row

Source: https://neuralcosmology.com/en/experiments/ba-ch01-qubit

An ideal qubit, with no noise or drift. Fourteen identical outcomes in a row at θ = 90° is one chance in 16,384: the “one in sixteen thousand” from the chapter.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/ba-ch01-qubit
From the book *Bugs Academy*, chapter [Peripheral Vision](https://neuralcosmology.com/en/read/bugs-academy/ch01.md).

> — There. Where a physicist publishes a paper claiming qubits respond to his intentions. Where reviewers laugh so hard their teeth fall out. Where a year from now we’re working in a high school, explaining Ohm’s Law to sixteen-year-olds.
>
> It came out convincing and terrifying — and beside the point, because I wasn’t talking about a publication. I was talking about the fact that my data was behaving the same way my life does: showing me something impossible, then pretending nothing happened. But I didn’t say that. I said:
>
> — Forget it. We’ll rerun the series.


---

# The Wow! signal: decode 6EQUJ5 yourself

Source: https://neuralcosmology.com/en/experiments/ea-ch08-wow

A printout from the Big Ear radio telescope (Ohio State University), 15 August 1977. The 6EQUJ5 column is genuine; the surrounding digits here are generated.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/ea-ch08-wow
From the book *Era of Architects*, chapter [Iero Returns](https://neuralcosmology.com/en/read/era-of-architects/ch08.md).

> (Radio astronomy, in this sense, has its own one-shot signal that was never explained. On August 15, 1977, the Big Ear telescope in Ohio caught a seventy-two-second pulse from the constellation Sagittarius, thirty times above background, hitting precisely the hydrogen line — in theory the most economical frequency for interstellar communication. The astronomer on duty, Jerry Ehman, circled the column on the printout in red pen and wrote in the margin: “Wow!” That has been its name ever since. There were no repeats over fifty-six years of subsequent observation. There is no explanation. …
>
> (e) Lia asked over lunch: “Papa, are you watching TV today?” I said: “Yes, once.” She thought. Said: “I’ll be in my room with Bradbury and Schrödinger Junior. Don’t come in the room until ten.” I nodded. Kira nodded. Lia went to her room. Kira and I looked at each other. Kira said one phrase: “She knows.” I didn’t answer — there was nothing to.

## What the Wow! signal is

On 15 August 1977 Ohio State University's Big Ear radio telescope picked up a narrowband signal from the direction of Sagittarius. A few days later the volunteer astronomer Jerry Ehman, going through the printouts, circled the column “6EQUJ5” in red pen and wrote “Wow!” in the margin. The name stuck.

The letters and digits on the printout are the signal's strength relative to background noise. The digits 0 to 9 stand for themselves, then come letters: A is 10, B is 11, and so on up to Z at 35. The sequence 6, E, Q, U, J, 5 shows the signal rising and falling as the telescope, fixed relative to the Earth, swept past the source. The peak “U” means the signal was about thirty times stronger than the noise.

The signal lasted about 72 seconds, exactly as long as a source stays in the telescope's beam, and arrived near 1420 MHz. That is the emission frequency of neutral hydrogen, long considered a natural channel for interstellar communication, which is why the signal became the most famous SETI candidate.

Neither Big Ear nor any other telescope has detected it again, despite decades of searching. A comet explanation proposed in 2017 did not hold up. In 2024 Abel Méndez's group proposed a natural cause: a cloud of cold hydrogen briefly brightened when a powerful transient source, such as a magnetar flare, lit it up. For now it is a hypothesis. In this experiment you can decode the real 6EQUJ5 column yourself.

## FAQ

### What does 6EQUJ5 mean?

The signal's strength relative to noise in six consecutive measurements. Digits stand for themselves and letters continue the count: A is 10, Z is 35. The peak “U” corresponds to about thirty times the noise level.

### Was the Wow! signal from aliens?

Nobody knows. The signal never repeated, so an artificial origin cannot be tested. There are natural hypotheses; the latest, from 2024, links it to a hydrogen cloud that briefly brightened.

### Was the Wow! signal ever detected again?

No. Big Ear, the Very Large Array, the Allen Telescope Array and others searched that part of the sky and found nothing like it.
- [Simulation theory: are we living in a simulation, or is the universe computing itself?](https://neuralcosmology.com/en/answers/is-the-universe-a-simulation)


---

# Karl’s module

Source: https://neuralcosmology.com/en/experiments/ea-ch10-synth

Karl’s module is invented; the sound is real: five detuned oscillators, a resonant filter and slow modulation, all built in Web Audio right in your browser.

Interactive experiment (runs in the browser): https://neuralcosmology.com/en/experiments/ea-ch10-synth
From the book *Era of Architects*, chapter [Alpha Partners](https://neuralcosmology.com/en/read/era-of-architects/ch10.md).

> (The idea — an instrument that mirrors its listener — isn’t original to Karl. Keith Fullerton Whitman, an American composer, since the early 2000s has been writing generative compositions on modular synthesizers: he launches a patch and sits for forty minutes without intervening, while the patch plays itself. With Whitman, the listener is Whitman himself. Karl wants to turn this idea a full revolution: the patch should play not itself but whoever is standing across from it. Technically, that means feedback by signature. Whitman knows nothing about signature, and so his generative runs evenly. …
>
> Karl plans today to try a five-minute regime. Olga, when she informed me of this at the gate, added one phrase: “if it starts to smoke — get the children out.” I noted it.


---

# Fluency on the cheap: epistemic hygiene in the LLM era

Source: https://neuralcosmology.com/en/essays/epistemic-hygiene

By Mikhail Savchenko, 2026-06-15

A practical problem arrived in 2023. Until then, writing a paragraph that looked like a specialist working through an argument required someone to think. The structure, the vocabulary, the hierarchy of claims, the references — all assembled in the head of whoever stood behind the text. Since 2023 the assembly has been done by machines, and done well enough that telling machine prose from human expert writing now costs separate effort.

Language models are a powerful tool, and I use them daily for code review, translation, literature search and quick first drafts of an argument. What worries me is public speech, where a smooth surface that resembles expert writing is increasingly taken for knowledge.

There's an old word for it that travels poorly across languages but the image survives. Picture a small-time reseller who has talked his way into selling a thermonuclear reactor out of a yard sale. The reactor is real, the labels are right, the leaflet is well written, and the reseller isn't lying. He simply doesn't understand what's in front of him and can't answer a single second-order question about it. Until recently this class of figure was rare at the upper levels of public conversation, because the entry threshold filtered them out. Now the threshold is a twenty-dollar-a-month subscription.

## Symptoms

A paragraph written by a machine and a paragraph written by a human who leans on a machine without doing any thinking of their own read the same. Here are the signs.

— The author risks nothing. Any substantive claim in science has the shape "here is a condition under which I am wrong." The yard-sale text never carries that condition, because the reseller doesn't know the condition under which his thesis fails.

— Smooth completeness. Real reasoning stumbles, retreats, catches its own counterexamples. The yard-sale text glides — it's generated as a stylistically consistent surface, and nothing past surface consistency survives a push.

— Names doing no work. In a serious text, a mention of Friston or Tononi or Levin either rests on a substantive connection (here is what is in their work, here is how that bends the argument) or it doesn't belong. In the yard-sale text, names function as "I'm familiar with the field" signals, and the actual work of the cited author plays no role in the argument.

— No numbers. Any claim about reality has to land somewhere in a number: an effect size, a power level, a range, a timescale. Yard-sale text routes around numbers, because numbers are where it could be checked.

## What sits under this

The "can a human tell ChatGPT prose from their own" test has mostly failed in its current form: the prose is good, manual fact-checking is slow, editors tire. The only filter that scales is the author, who has to filter their own position on the subject before sitting down to write.

Epistemic hygiene in this sense is a simple discipline. Before any public claim, ask four questions.

1. **Under what condition would I be wrong?** If the answer is a concrete observable outcome, keep going. If there is no answer, retract the claim.

2. **Where does this land in a number?** Order of magnitude is enough: "10⁻⁴ of effect width," "10⁵ trials," "10² galaxies." Without a number the claim is rhetoric.

3. **What is the strongest counterargument against me?** The most technical one an actual specialist would bring, with no strawmen. If you can't state it at the level the specialist would, you don't understand the subject.

4. **What would make the counterargument beat my thesis?** If nothing could, the thesis is theological. If something could, name it now, in advance.

The four questions aren't for every text; there is no point testing fiction or a personal essay this way. But any text that claims epistemic content — "here is how reality works," "here is what the experiment shows," "here is what follows from the theory" — needs to pass this filter before publication.

## Why I'm writing this to myself

I am writing this above all for myself, and for the version of me a year from now who will lose discipline and start publishing smooth, unanchored paragraphs about consciousness and physics. When that happens, this essay will be the checkpoint: under what condition was I wrong? where is the number? what is the strongest objection? what would make it beat me?

The same logic explains the shape of this site. The preprint ships as working code that anyone can rerun, with its falsifiers written into the manuscript — the conditions under which the programme dies. The essay on PEAR/GCP describes the protocol that puts it under fire. The essay on Levin's bioelectric memory issues a biological prediction that resolves or breaks in a single lab cycle. All of this is self-checking infrastructure built outside the author's head, because the inside-the-head check is the first one to weaken.

## What I ask of the reader

One thing. When you read a text that claims substantive knowledge of the world — mine or anyone else's — put at least the first question to the author. *Under what condition would you be wrong?* If there is no answer, the text carries no weight, however persuasive, elegant or well cited it may be. It is yard-sale fusion.

Surface literacy has become cheap, and the filter has moved inwards. Without it the public conversation about reality will stop meaning anything in short order. With it there is still a chance to tell a programme that can be checked from a piece of well-written persuasion.


---

# Levin's hologram: what bioelectricity says to distributed systems

Source: https://neuralcosmology.com/en/essays/levins-hologram

By Mikhail Savchenko, 2026-06-01

A simple, uncomfortable experimental setup from Michael Levin's lab at Tufts. A planarian is a freshwater flatworm about a centimetre long. Cut its head off, and within a week the head grows back. People have been doing this since the nineteenth century; nothing surprising so far.

If, however, you impose a specific pattern of membrane potentials on cells in the wound region during regeneration (Levin's lab does this with electrical and pharmacological tools), the planarian regrows *two* heads instead of one. That the voltage gradient steers the form is not a surprise.

But cut the two-headed worm in half *without any further manipulation*, and both halves regrow two heads again; the next round of cuts gives the same result. The two-headed morphology reproduces itself through multiple regeneration cycles with no genetic change and no continuing external stimulation. Durant, Morokuma, Fields, Williams, Adams, and Levin pinned the effect down in 2017 (*Biophysical Journal* 112, 10).

That means a part of the "body plan" is stored somewhere other than DNA — the DNA of both fragments is identical — and somewhere other than local protein assemblies, because after full regeneration the tissues are all new. Somewhere between the cells there is a persistent informational pattern that reconstructs the same morphology from whatever material gets pulled into its "field."

Levin calls this bioelectric memory. Technically, it consists of stationary patterns of membrane potential in the networks through which cells communicate. Phenomenologically, it is a hologram of the body plan spread across living tissue.

## What Pointer Architecture predicts

If an organism is organised as a distributed process with local terminals (cells) and a shared pattern memory (the bioelectric field), then a cell is a node reading values from an addressable structure rather than an autonomous computer.

Two testable predictions follow.

First, morphological memory should not reduce to local cell-cell signalling. If the memory is addressable, scraping the tissue out of a region doesn't destroy it — newly grown cells re-read it. This is already observed: the two-headed planarian survives amputation.

Second, and new: if the pattern is addressable, it ought to be *transferable* between organisms without transferring genetic material. Copy the electrical landscape of a two-headed planarian, impose it on an ordinary one, and the ordinary one should begin to build two-headed morphology on its next regeneration cycle. Technically hard, but possible in principle: voltage-clamp arrays and optogenetic control of ion channels supply the right toolkit. In 2012, Pai and colleagues showed the first step in *Development* (139, 313): by imposing an "eye" pattern on a Xenopus embryo's tail, they got an ectopic eye there, complete with retinal and lens cells.

## Xenobots: building a body with no blueprint

Kriegman, Blackiston, Levin, and Bongard in 2020 published in *PNAS* (117, 4) the Xenobots — self-organising biological objects assembled by an evolutionary algorithm from skin and cardiac muscle cells of the frog *Xenopus*, with the cells' genome left untouched and no externally specified plan imposed. Clusters of a few thousand cells exhibit coordinated behaviour — directed motion, aggregation, simple self-repair.

Classical biology has no explanation for this. The genome of these cells has never "seen" the Xenobot regime — the cells come from an embryo where their developmental lineage ends in skin epithelium. They coordinate behaviour on the scale of hundreds of microns anyway.

If Pointer Architecture is right, cells have the ability to query form from a distributed structure even when the genome doesn't encode such forms. Local configuration — position, neighbours, ion gradients — becomes an address; the response is read back from it.

The prediction here is sharp: with identical genomic and tissue inputs, varying initial bioelectric conditions should produce predictably different final morphologies. This can be tested with the existing xenobot protocols by varying membrane potentials systematically and recording the outcome.

## What this changes for the programme

The Pointer Architecture programme rests on three legs. The formal one is already built: the preprint builds the substrate as working code, with falsifiers stated in advance. The observational test from the attention essay is waiting on a preregistered RNG protocol.

The biological test is the most expensive and the strongest. Bioelectric memory of form remains a phenomenon for which modern biology has no agreed mechanistic theory. Pointer architecture issues a clean prediction here: the memory is addressable, therefore transferable independently of substrate. If it transfers, biology acquires a new vocabulary in which the cell is a client, not a server, of the morphological plan. If it doesn't, Pointer Architecture loses one of its three legs.

This experiment is not a one-person job; it needs a lab at Levin's level or one of equivalent technical reach. The essay exists so that the prediction is on the public record *before* the experiment happens. That is preregistration at the level of theory, and it matters for the same reason ordinary preregistration matters: so that nobody, the author included, retrofits a found result with "this is exactly what I predicted."

The companion book, *The Celestial Code*, walks through all three legs of the programme in one text. This essay is about the most uncomfortable of them.


---

# From substrate to PII — falsifiers in research and in regulated AI

Source: https://neuralcosmology.com/en/essays/falsifiers-in-research-and-regulated-ai

By Mikhail Savchenko, 2026-05-19

I write a preprint on the side. I also clean up AI systems for a living. People sometimes ask whether those two things touch each other — whether sitting with a 646-assertion test suite for a discrete substrate and configuring an audit trail for a multi-tenant LLM stack belong to the same person, or only to the same calendar. They belong to the same person. There is a discipline that runs through both, and once you see it the two activities stop looking like two careers and start looking like two surfaces of one habit.

The habit is: name in advance the condition under which you would have to throw your work away.

In the [Pointer Architecture preprint](https://neuralcosmology.com/en/science/pointer-architecture) this looks like a list of falsifiers, written into the body of the manuscript where nobody has to dig for them. If any substrate demonstration fails its assertions on the released code, the operational claim is gone. If area scaling breaks on other substrate topologies, the dark-energy result is an artefact of the lattice. If the consciousness measure fails to discriminate any of its five predictions, the framework falls back to the bare substrate. Each is a tripwire. Trip it, and the work I just spent two years on becomes a curio.

In a vendor security review for a startup that ships an LLM-backed product, the same habit looks different but is built from the same material. Before I write a single line of architecture I ask the team to name the condition that would prove the privacy claim wrong. What event, observable, logged, recoverable from the audit trail, would prove that the system is not doing what we say it is doing on the marketing page. If the team cannot name that event — cannot say "if `X` ever happens, our PII handling story is a lie" — the system does not have a privacy claim. It has a privacy mood.

Most AI products in regulated industries have a privacy mood.

They have an architecture diagram that points data flow at the right boxes. They have a vendor questionnaire that says yes in the right places. They have a copy of someone else's DPIA. What they do not have is a sentence that ends with "if `X` ever happens, we are non-compliant." And so when `X` happens — and `X` always eventually happens, because `X` is exactly the kind of event the org has implicitly committed to never noticing — the team finds out at the same time as the regulator. Or the user. Or the journalist.

The verification harness that ships with the Pointer Architecture preprint is, structurally, an audit trail. One command, `make verify`, reruns every demonstration and ends with "artifact status: reproducible". Anyone who wants to refute the paper can re-run my work bit-for-bit, change what I claim is not load-bearing, and watch the conclusions either survive or crack. Compliance work is the same shape: events go into an immutable log, the system commits to what counts as a violation before any violation occurs, and the auditor's job is to re-run the claim. Without that loop, a privacy architecture is decoration.

Working on both sides clarifies what each side is for. The preprint side teaches that *the cheapest moment to specify a falsifier is before you have the result you want*. By the time you have the result, the result starts to lobby for itself. You begin to feel that an exception is appropriate, that this dataset is unusually noisy, that the rejection criterion was always provisional. So you write the criterion down first, in plain language, and you publish it. Public commitment changes the cost function. The same trick is what makes regulatory immunity hold under pressure. By the time PII is leaving the building, no one in the room wants to define the leak as a leak. So you define it months earlier, with the regulator's language, in a document the engineers cannot edit silently.

It also runs the other way. The applied work teaches what science forgets when it stays in its journal habitat. A falsifier that nobody can act on is decorative. A reproducibility pipeline whose data lineage is opaque does not survive a hostile audit. Audit trails are useful exactly to the extent that they survive being read in bad faith by someone whose job is to find a gap. That is also the standard for a published model: it survives being read in bad faith by a reviewer whose job is to find a gap. The two crafts calibrate to the same thing, adversarial replication, whether or not they admit it.

So when I publish a preprint about a discrete substrate and then leave the desk to write up a PII flow for a fintech LLM stack, I am running the same epistemic instrument, just on a different surface. The preprint is the version where the adversary is the field. The audit trail is the version where the adversary is the regulator and the version of yourself who will lie to you, in eighteen months, when convenient. Both versions need the falsifier written down in ink before the work starts.

It is also why I keep both shops open. The research arm exists at [neuralcosmology.com](https://neuralcosmology.com), which is where you are now. The practice arm — where the same discipline lands inside other people's AI systems — exists at [mikefluff.com](https://www.mikefluff.com). For consulting on AI privacy and compliance architecture, [Regulatory Immunity](https://www.mikefluff.com/#services) is the relevant doorway; for the books and the preprint, [the science index](https://neuralcosmology.com/en/science) and [the essays](https://neuralcosmology.com/en/essays) here.

The thing the two share is older than either site. It is the rule that the most important sentence in any technical document is the one that says "and here is how this would be wrong."


---

# Superposition as `while(true)`: quantum mechanics through an engineer's eyes

Source: https://neuralcosmology.com/en/essays/superposition-as-while-true

By Mikhail Savchenko, 2026-05-03

Anyone who has ever run a generative loop knows the feeling. While the loop runs, every possible continuation *lives in it at the same time*. Step in — drop a `print`, take a `return`, read a value — and from the whole bundle of branches only one survives. The others simply do not get recorded as "what happened."

That picture sits so close to what physics calls wave function collapse that the question suggests itself: is it a metaphor, or a hint about how the world is built?

## What the formalism actually says

The Schrödinger equation evolves a state unitarily. That means that as long as nobody "looks," the system sits in a linear combination of possible outcomes with known amplitudes. Measurement is the only step that converts amplitudes into probabilities and lets one branch persist.

The equation does not describe that step; this is what is called *the measurement problem*. The Copenhagen interpretation (Bohr, Heisenberg) takes collapse as a primitive. Many-worlds (Everett, 1957) claims there is no collapse, only branching of the observer. Decoherence theory (Dieter Zeh in 1970, then Wojciech Zurek) gives a quantitative mechanism: interaction with the environment suppresses the off-diagonal entries of the density matrix and effectively "selects" a measurement basis. To the question "what is left after," decoherence is silent; it explains *why we see definite states rather than their superpositions*.

An engineer will recognise the mechanics of distributed systems here. While a transaction is running, the state is undefined. The commit fixes one slice. All other branches that opened along the way are discarded and never reach the final database. Decoherence is the analogue of losing transactional isolation: an external observer "sees" one branch because the intermediate states have ceased to be reachable.

That is still metaphor. From here it gets more interesting.

## The delayed-choice quantum eraser

There is an experiment where the metaphor starts to crowd reality.

Scully and Drühl proposed the scheme in 1982. It uses pairs of entangled photons: one flies to the main screen, the second carries a "tag" — information about which slit the first went through. While the tag exists there is no interference at the main screen. *Erase* the tag and the interference returns. The erasure can be performed *after* the main photon has already hit the screen.

Kim and colleagues (*Phys. Rev. Lett.* 84, 1, 2000) built it in the lab. It has since been repeated in many variants, including on the superconducting qubits of cloud quantum computers, where the program itself decides whether to erase the tag after the main qubit has been "measured".

The experiment does not send information backward in time, and this is exactly where popular retellings go wrong. Erasing the tag *regroups* data already recorded: it changes the joint distribution between main screen and which-path channel, and interference appears in the subset selected by the new condition.

In machine-learning terms, measurement is the loss function and erasing the tag is a backward pass. Nothing arrives "from the future"; the pass recomputes joint statistics over a different dependency graph. The forward pass already happened; its raw outputs do not change. What changes is *how we group them*.

That is the engineering content of collapse. A quantum system behaves as a distributed statistical structure in which the choice of measurement determines which joint distributions are accessible to us. The image of "reality waiting for an observer" only gets in the way.

## "It from bit": a blueprint instead of a slogan

John Wheeler in 1989 wrote down the line "it from bit": every physical quantity, at bottom, derives its value from binary yes/no answers — that is, from informational events. For years it sounded like a philosophical flourish without technical content.

Holographic codes (Pastawski, Yoshida, Harlow, Preskill, 2015 — the so-called HaPPY code) gave the first concrete model: spacetime as a network of tensors, with "deep" degrees of freedom reconstructible from "boundary" ones via specified unitary rules. If that picture is right, the physics of bulk objects is literally operations on a graph of pointers: add a node, rewrite an edge, repackage a codeword.

Holographic codes are a concrete mathematical object on which predictions can be tested: about reconstruction, about error tolerance, about entropy relations. Talk of resemblance ends here and engineering physics begins. The question is no longer whether the universe is "like" a computer: *part of its formalism coincides with the formalism of a specific class of computations.*

## Where the metaphor stops

Your `while(true)` does not, of course, literally generate quantum states. Decoherence on an ordinary chip is instantaneous because the environment is rigidly coupled to it through heat exchange. To preserve coherence long enough for a measurable quantum effect you need cryogenic conditions, isolation, specially engineered degrees of freedom — exactly what quantum-computing labs spend their lives on.

And no metaphor turns measurement into magic. Bell experiments (Delft, Vienna and NIST, all in 2015) showed that local realism is dead, which means a quantum correlation is far stranger than a "hidden parameter tucked inside the system". The generative-loop image helps catch the shape of that strangeness, but it cannot stand in for the physics.

## Why this parallel is worth drawing

The engineering view is useful because it makes quantum mechanics *discussable* for people who already have a working intuition about distributed systems, transactions and gradient descent; it does not pretend to explain the whole subject. There are millions of such people now, and for foundational physics that is a huge new audience.

The Pointer Architecture programme behind this site works right at that seam. Its claim is narrow and testable: *certain formal structures describe computation and physics at once, and testable predictions follow from that*. There is no "everything is code" slogan in it.

The preprint builds that loop into a working substrate, the Sixth language. The companion book, *The Celestial Code*, walks the intuition from `while(true)` to information geometry. This essay is about the cheapest bridge between the two languages. Anyone who wants to argue seriously about how reality is built should be willing to walk that bridge in both directions.


---

# Attention as load: what mass-observation experiments would mean if they hold up

Source: https://neuralcosmology.com/en/essays/attention-as-load

By Mikhail Savchenko, 2026-05-02

A small observation from an old and uncomfortable literature. Place a hardware random number generator next to a person concentrating on a particular outcome and the output statistics shift toward that outcome by something on the order of one part in ten thousand, an effect that separates from noise *only* over a very large number of trials.

Robert Jahn's Princeton Engineering Anomalies Research lab collected this kind of data from 1979 to 2007, several million sessions in all. The mean per-session shift was around 0.2σ, which compounds to roughly five sigma across the corpus. The lab closed with a plain statement: the effect is robust, the mechanism is unknown, the academic community did not accept the methodology. They never claimed a mechanism.

The paradigm continued in the Global Consciousness Project (Roger Nelson, 1998–present). A network of about seventy hardware RNGs scattered around the world records continuously. When something happens that draws mass attention — September 11, an Olympic opening ceremony, the death of a public figure — Nelson asks whether the aggregate statistic deviates from expectation. More than five hundred events have been published, with a composite z around 6.4σ.

This is uncomfortable for two reasons. First, the effect looks like a violation of independence in physical processes across a geographically distributed network. Second, the standard criticisms — file-drawer effect, post-hoc selection, flexible event definition — are *fair*. They don't close the question, but they reduce confidence in the PEAR/GCP meta-analyses to the point where no conservative physicist will cite them.

There is no need to take the data on faith. It is enough to notice the *shape* of the prediction that follows from them.

## What Pointer Architecture predicts

If reality is organised as a distributed computational system with local nodes and shared state, then like every such system it has bounded coordination bandwidth. When many observers query the same part of state at once — say, watching a single broadcast — the synchronisation load on that region rises.

What should show up under load? The same thing that shows up in any distributed architecture: rising latency, small desynchronisations, local deviations from stationary statistics. Mass consciousness does not move particles in this picture. During peak load the system simply maintains the local stationary process *slightly less well*.

The effect size in this picture has to be very small and very persistent. That is what PEAR and GCP find, and it is also what is hardest to tell apart from selection artefacts without a preregistered protocol.

## What it would take to confirm this or kill it honestly

One experiment is enough, preregistered on OSF. It fixes in advance a list of thirty upcoming global-scale events, each with an attention measure such as the forecast broadcast audience. It fixes the time window: ninety minutes either side of the peak. It fixes the statistic: a composite z-score across at least fifty hardware RNGs spread around the world. And it fixes the threshold: at the effect size GCP has claimed, the aggregate over thirty events must exceed 4σ.

If the threshold is met, you have a serious argument that synchronisation in physical systems depends on something that correlates with mass attention. If it isn't, twenty years of PEAR/GCP literature gets a clean "probably file-drawer."

Either way, we know more than we do now.

## Why bother with this essay

Within the Pointer Architecture programme, the formal half is already on testable ground: the preprint implements the substrate as working code, with falsifiers stated in advance. The biological half needs collaboration with labs of Michael Levin's calibre, which is not something one person can move alone.

Between the two sits an observational test that should be cheaper: equipment that already exists in a distributed network, and a protocol that fits inside one mid-sized grant.

Whether the effect is there, I do not know. What I do know is that it *ought to be* there if Pointer Architecture is right, that we have the instruments to check, and that nobody has yet run the check. That gap is easy to close.

The companion book, *The Celestial Code*, walks through the full prediction set. The preprint gives the formal model. This essay is about the cheapest of the tests, the one nobody has yet run properly.


---

# A Loss Function for the Universe

Source: https://neuralcosmology.com/en/essays/a-loss-function-for-the-universe

By Mikhail Savchenko, 2026-03-14

Here is a small experiment. Line up five of the most stubborn anomalies in contemporary science side by side.

The first is **galactic rotation**. Stars at the edges of spiral galaxies move faster than Newton says they should, given the mass we can see. The standard fix is to postulate large amounts of invisible matter, dark matter, distributed in halos around each galaxy. The fix works, in the sense that models with a well-chosen halo reproduce the curves, but no direct detection has arrived to end the argument.

The second is the **matter–antimatter asymmetry**. The Standard Model predicts that the Big Bang should have produced equal amounts of both; the universe we live in is made of matter, and the small amount of antimatter we detect is consistent with cosmic rays and collider physics, not primordial asymmetry. There are several proposed mechanisms — leptogenesis, electroweak baryogenesis — but none have been confirmed.

The third is **the measurement problem in quantum mechanics**. The Schrödinger equation evolves states unitarily until we measure them, at which point something happens — wave function collapse, or an effective collapse — that is not itself part of the equation. Interpretations multiply; the experimental record keeps refusing to single one out.

The fourth is **consciousness**. There is no accepted account of why a particular configuration of physical matter gives rise to subjective experience. Integrated Information Theory, Global Workspace, Predictive Processing — all plausible, none decisive.

The fifth is **cellular bioelectricity**. In the last decade it has become clear that cells compute collectively via bioelectric gradients in ways that look uncomfortably like the kind of coordination you would expect in a distributed system with shared state. Michael Levin's lab has kept producing results that are hard to reconcile with purely local chemistry.

These five live in five different journals. The usual thing to do with a list like this is to take each in turn and let the specialists fight it out.

I want to try a different move and ask what happens if we **fit all five with one loss function**.

## Why a loss function is a different kind of object

In machine learning, a loss function is the thing you minimise. It takes the current state of a system and says: here is how far from the target we are. Gradient descent is the procedure that nudges the system in the direction that reduces the loss.

Physics already uses objects like this — the action, the free energy, the variational principle. But there is a subtle difference. In ML, the loss is *defined by what we want*. The system doesn't know the loss exists; we impose it from outside. In physics, the analogous principle is that the system *is already minimising something* — action, entropy, free energy — and the universe falls out as the trajectory that does the minimising.

What if the five anomalies above share one loss function, and the reason they look separate is that we are looking at different coordinates of the same descent?

*Pointer Architecture*, the preprint of the research programme behind this site, takes this move seriously. The hypothesis is testable: if there is a single underlying minimisation, then departures from the accepted models at galactic scales should be linked to departures at biological scales. The preprint builds the substrate such a minimisation would run on as working code, the Sixth language, and states its falsifiers in advance.

You do not have to buy the hypothesis to notice that it *pays rent*. If it is wrong, it is wrong in specific, measurable ways. If it is right, it rewrites a lot of textbooks.

## What to measure next

I see three directions.

1. **A working substrate for the minimisation.** Already in the preprint: the Sixth language, 40 demonstrations, 646 automated assertions, no failures.
2. **Bioelectric coordination under perturbation consistent with a shared minimum.** Hard to do without collaborators; the Levin-lab techniques are the right starting point.
3. **Conditions under which the measurement problem looks different.** Quantum Zeno-like setups tuned to the predicted regime; this is where the experimental cost is highest.

Each direction is a bet, placed *publicly and in advance*. That is the only way to tell a scientific claim from a plausible story.

The companion book, *The Celestial Code*, walks through the argument in full. The preprint gives the numbers. This essay is the short version I wish I had read before writing any of it.


---

# Shifting the rhythm shifts what's available

Source: https://neuralcosmology.com/en/essays/frequency-hack

By Mikhail Savchenko, 2025-04-09

"Change your frequency and the world changes" is an esoteric slogan, but dig past the metaphor and you find an ordinary neurobiological fact. Your inner state really does shift what you see and respond to, because the brain works as a filter with adjustable settings and everything happens inside it.

Here is how it works.

## Attentional set: what it is and why it works

At any moment your perceptual field carries thousands of signals: visual, auditory, bodily. You become aware of a tiny fraction of them; the rest is screened out before it ever turns into experience.

Two things decide what gets screened out: biological priority (hunger, danger, motion) and the *current task* the brain is busy with. Psychologists call this the attentional set, and it works in every direction. Tune yourself to find a familiar face in a crowd and you will find it; tune yourself for grievance and you will find a cause; tune yourself for possibility and you will find that too.

Forty years and thousands of experiments stand behind this. The famous "gorilla in the middle of the basketball game" (Simons & Chabris, 1999) is exactly this effect: you miss a person in a gorilla suit because you were asked to count passes.

## Neural correlates

The brain generates different rhythms in different states:

— **theta** (4–8 Hz): a plastic state, good for learning;
— **alpha** (8–13 Hz): relaxed wakefulness, screening out sensory noise;
— **beta** (13–30 Hz): active problem-solving, narrow focus;
— **gamma** (30–100 Hz): binding perception together, unexpected connections.

When the brain switches between these modes, different neural ensembles come online and go offline, and esoteric "vibrations" have nothing to do with it. Meditation shifts the rhythms toward alpha and theta, anxiety toward high beta, and all of it shows up plainly on an EEG.

The world stays the same throughout. What changes is the filter, and with it what you *see* in the world.

## Three modes, three maps of the world

**Anxious mode (high beta, little alpha).** You see threats, failures, injustice and reasons to be angry; you make fast, narrow decisions; conversations fall apart, small things grate, the future looks dark. Biologically this is survival mode, useful over minutes and destructive if you stay in it for years.

**Calm mode (plenty of alpha, moderate beta).** You see the situation whole, notice nuance, hold several options in mind at once; conflicts untangle more easily, and the future looks *varied* rather than catastrophic. For a healthy nervous system this is the default mode, and it is the one modern environments knock out most often.

**Creative mode (theta with short bursts of gamma).** You see connections between things you never linked before, find unexpected solutions, write things you later cannot explain how you came up with. Biologically this is the mode of the hippocampus and the default-mode network, and it rarely survives the push for productivity.

## Practical takeaway

"Frequency hack" is good marketing and a poor description. Behind it lies a useful truth: you *do* have levers over your own perceptual filter, and they have been known for a long time.

— Physiology (breathing in for four counts and out for eight, cold, physical exertion) shifts the mode within minutes.
— Environment (the people you spend time with, what you read and watch, how much silence you get) shifts it within days.
— Meditation, long walks without a phone and keeping a journal shift it within weeks.

Each lever has its own time constant, and confusing them gets you nowhere: one meditation will not give you what a year of practice does, and a change of scene will not replace a breathing exercise.

## Where the line is

Everything above is standard neuroscience of attention and brain states, which is to say, work with your own nervous system. There is no "access to parallel realities" and no "quantum shifting" here.

If after a well-spent morning you feel that "the world is different", you are half right: the world stayed the same, and you changed. If you think about it, that is the only version of "reality shifting" with any practical use. The rest is marketing.


---

# How I arrived at Neural Cosmology

Source: https://neuralcosmology.com/en/essays/birth-of-neural-cosmology

By Mikhail Savchenko, 2025-03-04

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. How it got there, given that I am neither a neuroscientist nor a cosmologist, is the subject of the short timeline below.

## 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.

There was nothing metaphysical about it: anyone who has run an LSTM and a Transformer on the same task has seen the difference first-hand.

## 2020: Vanchurin's paper

In September 2020, Vitaly Vanchurin posted a preprint on arXiv titled *The World as a Neural Network*. The title sounded like a pop-science headline, but inside was a proper formalism. Vanchurin 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).

In 2022 Vanchurin, with Katsnelson and Koonin, published a follow-up in *PNAS*. Koonin is a well-known evolutionary biologist and Katsnelson a condensed-matter physicist; the journal is first-rank, and the work had nothing fringe about it.

It gave me an odd feeling. Someone had put into formal words what I had *felt* as an engineer but could not yet say as a scientist.

## 2020–2022: five cracks

After that, things piled up slowly. I started noticing related observations arriving from 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 kept developing IIT and began working it out for computable cases.

— New experiments gave Penrose and Hameroff some support: microtubules may hold coherence longer than previously thought.

The five news items came out in different journals, but laid side by side they make *one* 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 someone had to state this synthesis out loud, in Russian as well. In English it lived in fragments across academic Twitter and podcasts, and no book had yet collected it while keeping what is proven honestly apart from what is not.

I am neither a neuroscientist nor a cosmologist. But I had spent twenty years with systems where architecture determines what is possible, and I had talked with researchers from all five fields. It was an unusual position. A specialist has more depth and a journalist more breadth; what I had was an *engineer's view* of a science that had not yet become science.

## 2023–2025: Pointer Architecture

While I was writing the book, a working hypothesis of my own grew alongside it. I called it Pointer Architecture: a formal model of a discrete substrate in which everything else grows out of a first distinction and pointers, all the way up to an estimate of the dark-energy density.

In 2024 I assembled the preprint, and it has since gone through nine versions. The current one, v9.0, builds the substrate as working code — a small language called Sixth — with falsifiers written down in advance. The code is public.

This is a long way from proving Neural Cosmology. The preprint tests *one* of its consequences and says in advance which result would refute it.

## Where I am now

Neural Cosmology is a working hypothesis with clearly marked edges. Part of it is being tested, part is speculation for now, part is personal interpretation. I try not to mix those three.

If the hypothesis is right, science gains a new field. If it is wrong, what remains is a concrete, testable construction, and whoever takes up the question next will not have to start from zero.

Which way it goes, I do not know. That is why I keep going.


---

# Near-death experience: what neuroscience knows, and what it doesn't

Source: https://neuralcosmology.com/en/essays/dance-of-limits

By Mikhail Savchenko, 2024-12-18

I was born half-strangled and spent the first minutes of my life without oxygen. I remember none of it, of course. What I know about clinical death from other causes comes from several dozen reports I have read and cross-checked.

The near-death experience (NDE) is one of those phenomena where believers in the paranormal and materialists dug their trenches long ago and have been shooting at each other ever since. Hardly anyone on that field speaks in a calm voice. Let me try.

## What is actually observed

The classic NDE runs like this: leaving the body, moving through a tunnel, light, meeting relatives who have died, a *life review*, the return. Not everyone has one. It happens to a small but steady share of people who survive cardiac arrest or another brush with death: depending on the study, 9 to 18 percent of cardiac-arrest survivors report some form of NDE (van Lommel, 2001; Parnia et al., 2014).

The phenomenon itself is well documented. The trouble starts with its *interpretation*.

## What neuroscience says

Several neurobiological mechanisms are known today, and each explains some part of the NDE.

— **Hypoxia and reperfusion.** Starved of oxygen, neurons in the visual cortex begin to fire chaotically, and from the inside this looks like a tunnel with light at the end (Blackmore, 1993).

— **DMT and endogenous psychedelics.** Rick Strassman suggested that the dying brain releases its own DMT. This has not been confirmed in humans, but DMT has been found in the rat brain, and its level rises during cardiac arrest (Dean et al., 2019).

— **A gamma burst before death.** Borjigin and colleagues recorded a brief surge of gamma waves in the final seconds, first in rats (*PNAS*, 2013) and later in a few dying patients (*PNAS*, 2023), strongest in regions tied to perception and memory. The brain lights up one last time, brighter than in ordinary life. This may be what lies behind the "hyper-lucid awareness" so many survivors describe.

— **REM intrusion.** Nelson and colleagues (*Neurology*, 2006) found that people with NDEs more often have episodes in which REM sleep breaks into wakefulness. Hence the hypothesis that an NDE is a kind of dream set off under extreme conditions.

No single mechanism explains the NDE *as a whole*, but put together they account for most of what survivors report.

## What neuroscience does *not* know

This is where the line runs.

Reports of accurate details of the resuscitation, seemingly seen "from above" (the literature calls them veridical perceptions), are methodologically hard to work with. Pim van Lommel (*Lancet*, 2001) assembled a Dutch cohort of 344 patients; 18 percent reported NDEs, some with verifiable details. But the AWARE study (Parnia et al., 2014), designed specifically to test perception "from above", gave a modest result: of 101 patients interviewed, only one recalled an event that could be independently checked.

That result does not allow us to call veridical perceptions a myth. Their *rarity* and the *failure to reproduce them* do call for caution.

The long-term personality changes are also an open question. Many NDE survivors say their values shifted, their fear of death faded, their attitude to work changed. This can be tested statistically, and the tests confirm it. But it is explained well enough without any "contact with another reality": this is what post-traumatic growth looks like, a personality reorganising itself after an extreme experience.

## What I think

I sort my own conclusions into three levels of confidence.

**Established:** the NDE is a real neurocognitive phenomenon with a characteristic phenomenology. It occurs in 9–18 percent of cardiac-arrest survivors and changes personality for the long term.

**Preliminary:** part of the phenomenology (the tunnel, the light, the life review) is explained by known neurobiological mechanisms. The terminal gamma burst is the most promising lead on "hyper-lucid awareness".

**Speculation:** veridical perceptions, if they really exist, may point to something beyond the standard model of consciousness. Or they may be an artefact of memory, assembled after the fact from scraps of information. Right now the only *honest* answer is "I don't know".

In Neural Cosmology, the NDE is an interesting boundary case. If consciousness is information processing in a local node of a larger network, what happens when the node leaves the network? The standard answer: the process stops. The non-standard one: for a brief moment the node loses its "filter" and sees what is normally out of its reach.

Which answer is right, I do not know. And for as long as that "I don't know" stands, I keep the door ajar. Only ajar, so that no draught blows in from the office next door, where "contact with spirits" is sold for $150.


---

# Topology of mind: why Poincaré is a useful language for thought

Source: https://neuralcosmology.com/en/essays/topology-of-mind

By Mikhail Savchenko, 2024-11-04

The Poincaré conjecture is simple to state: every closed, simply connected three-dimensional manifold *is* a sphere, up to continuous deformation. For a hundred years nobody could prove it. In 2002 and 2003 Grigori Perelman posted three preprints to arXiv that settled the problem, then turned down both the Fields Medal and the million-dollar Clay prize and went on living in his flat in Kupchino, on the edge of St Petersburg.

Now for a few uncomfortable observations.

## A coherent thought behaves like a closed manifold

When a person thinks something through in sequence, building an argument, reaching a conclusion, returning to the start, their neural activity leaves a *trace* through state space. That trace can be recorded, and it is: in fMRI, in EEG, in neural embeddings.

Over the past decade topological data analysis has been gaining ground in computational neuroscience. The idea is simple: instead of treating brain activity as a vector in Euclidean space, you look at its *shape*, at connectivity, holes and loops, and at which of them persist as the scale changes.

A strange regularity surfaces. Coherent thought resembles activity that closes into a loop; anxious, scattered thought resembles activity whose ends never meet. Some studies suggest that in meditation this activity has fewer "holes" than in ordinary waking.

The Poincaré conjecture itself plays no direct role here. But its *language* turns out to fit what we observe, and that is already something.

## What it means

A few studies worth knowing.

— Giusti and colleagues (*PNAS*, 2015): topological analysis of correlations between hippocampal neurons reveals geometric structure that ordinary methods miss. In this picture geometry grows out of topology.

— Saggar and colleagues (*Nature Communications*, 2018): the Mapper algorithm turns fMRI data into a graph in which brain states during different tasks fall into distinguishable regions, and transitions between them can be tracked in a single person.

— Northoff and Huang (2017), the "temporo-spatial theory of consciousness": consciousness as a structure with characteristic dynamics in time and space.

These data are under no obligation to prove that consciousness *is*, in substance, a topological manifold. What they show is something else: *topology is a fitting language* for describing how consciousness works.

The distinction matters. Ontology answers "what is", methodology answers "how to describe", and so far we stand on the second.

## Why look at this at all

Computational neuroscience today is roughly where physics was at the end of the nineteenth century: plenty of data, no general model. Everyone knows that neurons fire and that their activity is coordinated. But *what* is coordinated, and *how*, is described at a level Rutherford would have called stamp collecting.

Topology is an attempt to give the future model a mathematical skeleton. If coherent thoughts are closed loops in state space, if lucid awareness corresponds to particular persistent homologies, if déjà vu is current activity crossing a previously recorded trajectory, then *thinking* becomes something you can define formally.

"Mathematics has explained consciousness" claims more than we have. The careful version is that mathematics offers a tool with which we may at last begin to measure consciousness.

## Where the evidence runs out

To prove the Poincaré conjecture, Perelman relied on the Ricci flow, which Richard Hamilton had been developing since the 1980s. The Ricci flow describes how the metric on a manifold changes under the influence of curvature.

Some papers try to apply the Ricci flow to graphs of neural connections. It is an elegant and fashionable move, but a premature one: a mathematical analogy becomes a theory only when it makes predictions that can be tested.

So far the topology of mind has produced language, metrics and hypotheses. It has not produced a single *prohibition*, not one "if this holds, that cannot be observed", and without prohibitions a hypothesis does not become science.

We are in the phase where language runs ahead of theory. That is how a field usually looks before a breakthrough, or before discovering that no breakthrough is coming. The second outcome is a useful result too.


---

# Quantum consciousness: what is proven and what isn't

Source: https://neuralcosmology.com/en/essays/quantum-consciousness

By Mikhail Savchenko, 2024-11-03

In 1935 Einstein, Podolsky and Rosen published a paper now known as the EPR paradox. Formally, the authors wanted to show that quantum mechanics is incomplete; in practice, they gave the first description of what would later be called quantum entanglement.

Einstein called it *spooky action at a distance*. He was convinced it was an artefact and that a deeper local theory with hidden variables stood behind quantum mechanics.

In 2015 Ronald Hanson's group at Delft University of Technology ran a loophole-free Bell test: electron spins in diamond defects 1.3 kilometres apart, with neither the locality loophole nor the detection loophole left open. The result: Bell's inequality is violated, and local realism is *dead*.

This measurement is no longer up for dispute. The same year, similar results came from Zeilinger's group in Vienna and from NIST in the United States, and later from Munich, all telling the same story.

## What this does *not* imply

Quantum nonlocality does not get you to a brain that is a quantum computer, to consciousness smeared across the universe by entanglement, or to a mind you can "upload" to the cloud and "tune" to a field.

The usual interpretive mistake is to take a real physical fact (entanglement exists) and bolt a speculation onto it (entanglement operates in the brain), as if the first supported the second. It does not.

## What this *may* imply

Since the mid-1990s Roger Penrose and Stuart Hameroff have been developing the Orch-OR hypothesis: coherent quantum states exist in neuronal microtubules, and their collapse is tied to moments of conscious choice. Most physicists were sceptical, since at body temperature decoherence should destroy any coherence almost instantly.

Recent years have given the hypothesis some support. In 2023 Gregory Scholes's group at Princeton showed that excitation energy travels along microtubules farther than classical models predict (*ACS Central Science*). In 2024 Philip Kurian's group at Howard University found superradiance in tryptophan networks inside microtubules, a collective quantum effect that persists at room temperature (*Journal of Physical Chemistry B*). This is a long way from proving Orch-OR, but the hypothesis can no longer be dismissed out of hand.

It is an interesting position: a theory that once sat outside serious discussion has moved into testable speculation.

## Tononi and integrated information theory

Giulio Tononi proposed a different approach, integrated information theory (IIT). Consciousness, for Tononi, is a measure of how *integrated* a system is, written Φ (phi). The more a system behaves as a single whole and the less as a set of independent parts, the higher its Φ.

The theory is elegant. Its main weakness is computational: the exact Φ of a system the size of a brain cannot be computed, because the complexity grows exponentially. Critics led by Scott Aaronson point to something else as well: IIT makes strange predictions for artificial systems, for instance assigning a simple grid of logic gates more consciousness than a human. Tononi acknowledges the problem but keeps the theory.

For an honest theory this is a normal position: it makes predictions that can be criticised.

## What I think

I sort my conclusions into three levels of confidence.

**Established:** quantum mechanics is nonlocal (Delft, 2015, and later experiments); information is physical (Landauer, 1961; experiment, 2012); consciousness depends on how integrated brain activity is, and the number of neurons alone decides nothing: the cerebellum holds most of them yet contributes almost nothing to consciousness.

**Preliminary:** microtubules may sustain quantum effects longer than previously thought. Orch-OR has become a subject of discussion, though it remains unproven.

**Speculation:** if gravity is emergent (Verlinde), the universe is a neural network (Vanchurin) and consciousness is integrated information (Tononi), then all three claims may describe one phenomenon from different sides. That is speculation, of course, but speculation you can start testing.

Quantum consciousness today is a set of *measurements*, each of which is still alive, with no finished theory underneath them yet.


---

# Echo of consciousness: time as a byproduct of information processing

Source: https://neuralcosmology.com/en/essays/echo-of-consciousness

By Mikhail Savchenko, 2024-11-01

John Archibald Wheeler, the physicist we owe the term "black hole" to, formulated in 1989 a principle that fits into three words: *it from bit*. Everything that exists comes from information. Matter, space and forces, if Wheeler is right, are all derivatives of a more fundamental layer in which *bits* live.

At the time it sounded like metaphor with ambition. Today it sounds like a research programme.

## Landauer: information is physical

In 1961 Rolf Landauer showed that erasing one bit of information must release *at least* kT·ln 2 joules of heat; going below that threshold would break the second law of thermodynamics. In 2012 Bérut and colleagues (*Nature*) measured it in the lab. They held a single colloidal particle in a double-well trap and erased the information about its position, and the slower the erasure, the closer the heat released came to Landauer's limit.

The result is enough to close the question. Information has a *thermodynamic cost*. It is physical in the same sense energy is.

## Delayed choice: the strangeness that stays strange

Delayed-choice experiments have been plentiful over the last twenty years. The basic scenario: a photon flies through an interferometer; the decision how to measure it (as a particle or as a wave) is made *after* it has already traversed the device. Classical intuition says this cannot work, because the outcome should already be fixed by what the photon has done.

The results agree with quantum mechanics: the photon behaves as the chosen measurement requires, even when the choice is made later. One interpretation: information travels backwards in time. Another: "what the photon did" before measurement has no meaning at all. Physicists prefer the second, but it too destroys the intuitive picture of time as a tape running left to right.

## What this means for consciousness

Two facts have to be reckoned with.

1. Information is a physical object with a thermodynamic cost (Landauer, Bérut).
2. Quantum events have no well-defined "before measurement" (Wheeler's thought experiment; the experiment by Jacques, Aspect and colleagues, 2007).

Combine these and you get a picture in which *time* is the way local information processors (brains, for example) *order* their interactions with the universe. The arrow of time is the gradient of entropy increase *for a given observer*. Outside a specific computational process, "when" has no physical content.

The idea is not mine. Carlo Rovelli wrote a book about it (*The Order of Time*, 2017). Julian Barbour did so earlier still (*The End of Time*, 1999). Barbour in fact thinks there is no time at all, only a Platonic space of instantaneous configurations of the universe, with "flow" an illusion. I do not go that far, but physicists take the position seriously, and brushing it aside would be dishonest.

## Where speculation starts

**Not speculation:** information is physical, time has been emergent in statistical mechanics since Boltzmann (1872), and quantum measurement has no "before".

**Half-speculation:** the brain is a local information processor, and its sense of time is a byproduct of its computations. Neuroscience supports this: *perceived time* stretches and shrinks depending on what the brain is busy with (Eagleman, *Current Opinion in Neurobiology*, 2008). Whether the subjective arrow of time matches the physical one is still unknown.

**Speculation:** "echo of consciousness", "déjà vu as information time-travel". The phrases are pretty, but no measurement yet separates them from the ordinary workings of memory.

I file this under "interesting to test"; "proven" is a long way off.

If Wheeler is right, consciousness is what *produces* time, locally, out of bits.


---

# Magic and reality: where science agrees and where it doesn't

Source: https://neuralcosmology.com/en/essays/magic-and-reality

By Mikhail Savchenko, 2024-10-19

Magical thinking is the belief that a thought, an intention or a symbol can act on the physical world while bypassing ordinary causal links. Anthropologists have studied it since the 1890s, psychologists for about as long. The short cultural verdict fits into three words, "archaic", "childish", "superstitious", but the scientific picture is longer and more complicated.

I will take it in three steps.

## Where magical thinking *works*

The placebo effect is a real physiological process. A meta-analysis by Hróbjartsson and Gøtzsche (*NEJM*, 2001) found it small, but that was for objective outcomes. For subjective experiences such as pain, nausea and anxiety, placebo works noticeably better. The mechanism has been traced: as early as 1978 Levine and colleagues showed that naloxone cancels placebo pain relief, and later Fabrizio Benedetti described in detail how expectation switches on specific neurotransmitter systems, including endogenous opioids and dopamine.

The upshot is simple: if you believe a remedy will help, your brain produces chemical changes that really *do* help. It is biochemistry triggered by a signal from above, from the higher cognitive levels. Magical thinking serves here as an *interface* that leads a person to a real mechanism, and the mechanism itself stays entirely material.

The same holds for prayer, meditation and ritual. Neuroimaging (Newberg and colleagues, 2001, and many studies since) shows measurable shifts in the prefrontal cortex and parietal regions during deep prayer. The neural network runs in a particular mode, and that has physiological consequences: anxiety drops, the pain threshold changes, sometimes the brain's structure changes over time. None of it requires an outside agent.

## Where magical thinking *misleads*

Confirmation bias makes us remember the hits and forget the misses: you wished for something and it happened, so you remember; you wished and it did not, so you forget. Daniel Kahneman described this in detail (*Thinking, Fast and Slow*, 2011), though the effect was known long before.

The brain stitches random events into patterns and gives them meaning where there is none. That is useful in evolutionary terms (better to mistake a bush for a tiger than the other way round), but in modern surroundings it keeps producing false "synchronicities".

Magical thinking about other people is more dangerous. "I thought something bad about him and something bad happened, so it is my fault." Or: "I sent him a blessing and he recovered, so I am a *miracle worker*." Both are ego traps in spiritual costume. Your thoughts connect to what happens to another person only through physical channels: you called, you said something, you did something.

## Where the boundary blurs

Quantum mechanics handed magical thinking an unexpected gift, the word "observer". Popularisers have been repeating the same mistake for twenty years: "if the observer affects a quantum measurement, then my consciousness creates reality".

The short physics answer is no. In quantum mechanics the "observer" is a *measuring apparatus*, and consciousness has nothing to do with it. Decoherence comes from the quantum system interacting with its environment; Wojciech Zurek showed this in detail in a series of papers in the 1990s and 2000s.

The longer answer is more complicated. Penrose and Hameroff argue that quantum coherence *inside the brain* may play a part in the moment of conscious choice. It is a hypothesis and only a hypothesis: there are fresh arguments in its favour (for example, a 2024 study in *eNeuro* in which a microtubule-stabilising drug delayed loss of consciousness in anaesthetised rats), it has not been accepted, and it has not been refuted. In its narrow form, "quantum processes in microtubules may be linked to consciousness", it is a testable claim. In its broad form, "consciousness creates reality", it is a slogan with no physical meaning.

## What I think

Magical thinking is the *forerunner* of scientific thinking. It says there is a hidden link between intention and the world. Science answers that some of these links are real and run through neuroendocrine and social mechanisms, some are illusions produced by cognitive biases, and some are unknown and remain an open question.

The sensible stance is to *separate*: what exactly you observe, what mechanism might stand behind it, and how it could be tested. No dismissive "superstition", and no esoteric add-on about "another reality".

It is duller than "the universe is listening". It also works.


---

# Whisper of stones: where consciousness ends

Source: https://neuralcosmology.com/en/essays/whisper-of-stones

By Mikhail Savchenko, 2024-09-04

If on a mountain walk you catch yourself thinking "these stones feel something", you are either tired or up against one of the oldest questions in philosophy: where does consciousness end?

The serious literature gives three answers. All three are workable, and none has won. Let me take them one by one.

## Answer one: nowhere (panpsychism)

Panpsychism holds that consciousness is a basic property of matter: it lies at the foundation and cannot be derived from how matter is organised. An atom, a molecule and a stone all have it, only in a homeopathic dose. The brain is built to *concentrate* this property locally into what we call experience.

David Chalmers (*The Conscious Mind*, 1996), Galen Strawson and Philip Goff take the position seriously. Their argument is formal: emergence does not explain *why* some physical processes come with experience and others do not, so it is simpler to assume that experience is a basic property of matter.

The criticism: panpsychism makes no testable predictions. What does it say about a stone? "It feels, but not the way you do." What does it forbid? Nothing. And a hypothesis without prohibitions is not science.

## Answer two: where integration ends (integrated information theory)

Tononi holds that consciousness is a property of *structure*, namely of integrated information, Φ. What matters is how the connections are organised, not what the system is made of. If Φ is above zero the system experiences something, and the higher Φ, the richer the experience.

Φ rises with integration: close to zero for an atom or a stone, a little higher for a microbe, much higher for a human, and still disputed for AI models.

The theory makes testable predictions, the main one being that integration strengthens consciousness and its breakdown extinguishes it. Clinical work bears this out: the perturbational complexity index tells wakefulness apart from sleep, anaesthesia and the vegetative state (Casali et al., *Science Translational Medicine*, 2013).

The theory's main problem is computational: the exact Φ of a system the size of a brain cannot be calculated. On top of that, it makes strange predictions for symmetric artificial systems: a lattice of logic gates in the right configuration gets a Φ above a human's. Either something is wrong with the theory, or we have not yet understood what it gets right.

## Answer three: consciousness as an interface (Hoffman)

Donald Hoffman (*The Case Against Reality*, 2019) argues that perception does *not* show us the world as it is. Evolution optimises survival and readily sacrifices truth for fitness. Our "world" is an interface simplified down to icons and buttons.

There is nothing esoteric about this. Using evolutionary game models, Hoffman and his co-authors showed that strategies tuned to fitness generally drive out strategies tuned to truth (Mark, Marion and Hoffman, 2010), and later proved it as the "fitness beats truth" theorem (Prakash and colleagues, 2021). So we are built to see a *convenient* picture, and accuracy is a bonus on top.

A curious consequence follows: "consciousness" as it appears to us may itself be part of the interface. We see ourselves as experiencing because that is efficient, and nobody promised us ontological accuracy.

Hoffman goes further and suggests that reality consists of "conscious agents" interacting through mathematical structures, with matter as an emergent description of those interactions. That is speculation, but it starts from a theorem.

## Where I stand

None of the three answers is settled. For working purposes I prefer integrated information theory, because it gives a parameter you can try to measure. Panpsychism stays interesting but offers no lever, and Hoffman is useful as a correction to naive realism.

A stone most likely feels nothing in the sense of human experience, and even an ant feels incomparably more. But the question "is a stone's Φ above zero?" is technically well posed, and it has to be answered by measurement rather than intuition.

If you sense a "presence" in a stone, that says more about your neural activity than about the stone. Still, the fact that your neural activity is capable of it is interesting in itself. Perhaps that is the real value of the walk: you entered a mode in which you would have heard the stone, had it anything to say.

A fine distinction, but an important one.


---

# Mirrors: where the fractal is real, and where it is a coincidence

Source: https://neuralcosmology.com/en/essays/mirrors-of-infinity

By Mikhail Savchenko, 2024-08-30

Fractal pictures are a worn-out staple of popular science. Bronchi look like a tree, a snowflake like coral, the brain's neural network like the cosmic web. It is beautiful, it goes on Instagram, and it gets captioned "everything is connected".

The trouble is that *looking similar* proves nothing. The brain is trained to find resemblance and finds it even where there is none; this is called pareidolia, and the classic example is a face on burnt toast.

Fractality, on the other hand, is a *mathematical property*, and it can be measured. That is where things get interesting.

## What is measurable

Fractal dimension describes how much complexity an object shows as you zoom in. A flat surface has dimension 2, a solid ball 3, and a fractal has a fractional value somewhere in between. In these terms the coastline of Britain comes out at about 1.25, the bronchial tree at about 2.9, the surface of the cerebral cortex at about 2.7.

These numbers are *testable*. They are found by box counting: cover the object with grids of different sizes and see how the number of non-empty boxes changes with scale. The result is an ordinary log-log plot, and the slope of its line is the dimension. No mysticism enters anywhere.

In 2020 Franco Vazza and Alberto Feletti compared two objects:

— a slice of cerebral cortex under a microscope;
— a computer simulation of the large-scale structure of the universe.

They looked at how density fluctuations are distributed across scales and at how the networks are built: how many connections each node has, how tightly nodes cluster. For several of these measures the numbers came out close, even though the objects differ in size by about twenty-seven orders of magnitude.

This is a peer-reviewed measurement, published in *Frontiers in Physics*.

## What it means and what it does not

**It means** that two systems differing in size by a one followed by twenty-seven zeros are similar in several *structural* characteristics: connections per node, the spectrum of fluctuations, the degree of clustering.

**It does not mean** that one system is a "reflection" of the other. A match in *mathematical structure* says nothing yet about a kinship in *substance*.

Bronchi and trees also have close fractal dimensions, yet bronchi are not tiny trees and trees are not giant bronchi. Both systems solve *the same problem*: get as much contact surface as possible for as little volume as possible. The same optimisation problem yields the same geometry.

Perhaps the brain and the cosmic web also solve a shared problem: connect as many nodes as possible while spending as little material as possible, and pass information along the shortest edges.

If so, the resemblance follows from a shared optimisation principle, and the mystical "everything is one" has nothing to do with it. What appears instead is a genuine *scientific* question: what exactly is the universe optimising, if it arrives at the same geometry as a brain honed by evolution?

## Where I stop

The brain and the universe are structurally similar under certain metrics. That fact deserves discussion. Calling them "the same thing" steps past what the data say.

The pretty phrase "each of us is a miniature copy of the universe" steps past it too. The data say something more modest: the *architecture of connections* in both systems obeys the same optimisation principles.

That is enough. Tacking on "so we are all neurons of God" adds no information and makes the information harder to hear: the listener either dismisses the whole thing or swallows it whole, "everything is one" included.

Better to say it precisely. The precise version is surprising enough on its own.


---

# Balance as a regime

Source: https://neuralcosmology.com/en/essays/dance-of-balance

By Mikhail Savchenko, 2024-08-29

In popular literature the word "balance" usually means you must balance work and rest, reason and feeling, masculine and feminine. It sounds nice and explains nothing. Balance is a *regime*: a mode the system enters, keeps falling out of and keeps returning to. You cannot stop inside it; you can only live in it.

The engineering name for that is feedback.

## Homeostasis: engineering inside the body

In 1865 Claude Bernard stated it clearly: the internal environment of the organism stays constant *despite* change outside. Today we know the numbers. Body temperature holds near 37 °C, give or take half a degree; blood pH stays between 7.35 and 7.45; glucose keeps to a corridor of 4–6 mmol/L.

The body does not "find balance" and then stand in it. Those numbers are held by dozens of feedback loops that measure deviations and correct them all the time: when temperature drops, shivering and vasoconstriction switch on, when it rises, sweating and vasodilation; when glucose climbs, insulin is released, when it falls, glucagon.

What you are looking at is a *system with a regulator*, with a setpoint, sensors and effectors, and "opposites in harmony" has nothing to do with it. When regulation breaks down, as in diabetes, no "inner wisdom" is held in reserve: insulin has to come from outside and blood sugar has to be watched from outside too.

## Cybernetics: the same for any system

Norbert Wiener (MIT, *Cybernetics*, 1948) showed that any system with negative feedback behaves the same way: it measures the deviation, applies a correction and returns to the setpoint. The same mathematics describes a thermostat, a heartbeat, a market price, and a political opposition acting as a stabiliser of power.

Balance, then, is described by the dynamics of return. The state of a system is set by how often it returns, how far it strays and how long it takes to react.

Feedback that is too fast makes a system oscillate, like a crane swinging on its cable; feedback that is too slow lets it drift, like a neglected political regime. Everything interesting happens in the narrow band between those extremes.

## What it means for a human

Two things follow.

**First, balance means return.** A system that never deviates is dead, because it has nothing to stabilise. Life is, by definition, a sequence of deviations and returns, and stability is measured by how well oscillations are damped.

**Second, different systems regulate at different speeds.** Body temperature is regulated over seconds, body weight over months, psychological state over anything from minutes to years, and confusing these speeds is the source of most mistakes. Trying to fix depression in a day works about as well as heating a room with a thermometer.

## Where the evidence runs out

In biology, homeostasis is a measurable engineering property. In psychology, "emotional balance" has neurological correlates (the prefrontal cortex modulates the limbic system), but a precise cybernetic model is only beginning to take shape. In social systems it is murkier still.

The principle is the same everywhere: feedback with the right time constant. Where it exists, there is stability; where it does not, sooner or later there is catastrophe.

"Balance of opposites" is a poetic way of saying "a stable regime with well-tuned feedback loops". The poetic version is easier to remember. The engineering version works.


---

# The web: what "everything is connected" actually means

Source: https://neuralcosmology.com/en/essays/web-of-the-universe

By Mikhail Savchenko, 2024-08-15

Saying "everything in the universe is connected" is about as useful as saying "everyone has a phone". It is true and completely useless, because the interesting part is always in the details: with whom, how often, over what channel, at what bandwidth.

Over the last twenty years science has given concrete answers to those questions. Let me go through them.

## The network of human ties is measurable

Social networks have been studied as graphs since the 1950s, but serious measurement began in the 2000s, when big data arrived. Nicholas Christakis and James Fowler (*NEJM*, *BMJ* and other journals, 2007–2009) showed that obesity, smoking, happiness and loneliness spread through a social network up to three degrees of separation: the friends of your friends' friends influence your weight. The effect is small but statistically robust.

How much of this is causal is still debated, but the channels, if they work, are entirely mundane: conversation, shared activities, imitation. No mysticism is needed. The network exists, it works, and it can be measured.

## Ecosystems are networks you can see

Robert May (Oxford, *Nature*, 1972) showed mathematically that an ecosystem that is too densely connected is, on the contrary, *less* stable: adding links does not by itself add stability. Connectedness can be productive or fragile, and the character of the links matters more than their number.

A tropical forest and a coral reef are both densely connected, yet the forest recovers after a fire while the reef barely recovers after bleaching. The difference lies in how the links are arranged: which nodes are key, where there is slack and where there is none.

The phrase "everything is connected" skips exactly this structure, though it is the only thing that makes systems interesting. Connectedness is a parameter, and it can be tuned.

## Quantum entanglement: a connection of a different kind

Entanglement is real (Delft, 2015) and nonlocal, but it cannot be used to send information. That is a proven result, the no-communication theorem.

In practice this means the following. Two entangled particles at opposite ends of the galaxy "know" about each other in a certain sense, but neither Alice nor Bob can use that "connection" to send a message. For that you need a classical channel: light, a signal, a courier.

So citing entanglement as proof that "we are all telepathically linked" misapplies the concept. There is a connection there, just of a very different kind from the one people want.

## What remains

Three layers of connectedness *actually* work.

1. **Social.** Behaviour, health and mood spread through real contacts up to three degrees out. This is measurable and useful in practice: it pays to choose your surroundings, because they filter what reaches you.

2. **Ecological.** Ecosystems are linked by food chains, chemical cycles and migrations. Knock out one node and a cascade runs through the network, most visibly when keystone species go extinct.

3. **Informational and physical.** If gravity is emergent (Verlinde) and the universe is a neural network (Vanchurin), then connectedness at the deepest level simply *is* the architecture of the universe. That is speculation, but testable speculation.

"Everything is connected" in the sense of "there is a path between any two points" is true but trivial. "Everything is connected" in the sense of "there is bandwidth between them and something flows" is far more interesting, provided you leave the magic at the door and count the graph.


---

# Resonance: physics term or esoteric slogan

Source: https://neuralcosmology.com/en/essays/symphony-of-cosmos

By Mikhail Savchenko, 2024-08-07

The word "resonance" has had a strange career. In physics it means something specific: an oscillating system amplifies its response to an external driver whose frequency matches its own natural frequency. That is measurable, and a great deal rests on it, from bridges brought down by resonance and a tuned radio receiver to MRI, which is built on nuclear magnetic resonance.

In esoterica "resonance" serves as a *synonym for "connection"* without specifying anything: "resonance with the universe", "resonance with another person", "raise your vibration". Such phrases are not necessarily false; they simply carry no information, because they do not say what is oscillating or at what frequency.

What interests me is the middle ground, the place where physics overlaps with what people call states of consciousness.

## What definitely exists: neural synchronisation

The brain generates electrical oscillations that can be picked up from the scalp with EEG. By frequency they are usually divided like this:

— **delta** (0.5–4 Hz): deep sleep;
— **theta** (4–8 Hz): drowsiness, meditation, learning;
— **alpha** (8–13 Hz): relaxed wakefulness;
— **beta** (13–30 Hz): active thinking;
— **gamma** (30–100 Hz): attention, binding perception together.

These rhythms are produced by ensembles of neurons firing in sync. Synchronisation is a physiological process: it can be shifted (by meditation, stimulation, drugs) and measured.

When two people interact, whether talking, touching or doing something together, *interpersonal synchronisation* appears: the rhythms of their brains begin to align. This is measured in experiments that record two brains at once, known as hyperscanning (Dumas et al., *PLoS ONE*, 2010; Dikker et al., *Current Biology*, 2017).

When people say "we're on the same wavelength", they are unknowingly describing a specific phenomenon: the alignment of rhythms and the timing of neural ensembles across two separate brains.

## What possibly exists: resonance with the environment

Schumann resonances are standing electromagnetic waves in the cavity between the Earth's surface and the ionosphere. The fundamental is about 7.83 Hz, with overtones near 14, 21 and 27 Hz. This is measured and undisputed.

Does the human brain resonate with these frequencies? Some researchers (König, Cherry, Persinger) say yes. Most of that work is methodologically weak, with small samples and no preregistration. In the serious literature the question remains open.

The same goes for geomagnetic fields. There are statistical correlations between geomagnetic activity and human behaviour (aggression, anxiety, strokes). The effects are small but measurable (Palmer et al., 2006), and the mechanism is unknown.

None of this counts as proof. It is an example of a grey zone: conscientious scientists work there with testable hypotheses, but there is not yet enough evidence either way.

## Where resonance definitely *fails*

"Resonating with another person's intention through walls, at a distance." This has been tested many times, under controlled conditions and with preregistration, and no effect has been found. Dean Radin of the Institute of Noetic Sciences claims otherwise, but his meta-analyses have been criticised on methodological grounds (Hyman, 2010), and they do not survive independent replication.

"Resonance with crystals, stones, pyramids", in the sense of drawing energy or information from them, is physically meaningless. A crystal does not emit electromagnetic oscillations of any significant power on its own; the quartz in a watch oscillates only because an electrical circuit drives it.

## What to take away

When someone says "resonance", ask: at what frequency, between which systems, and how is it measured?

If the answer is "two brains synchronising in conversation", that is a physical fact you can study. If it is "the brain and the Earth through Schumann frequencies", that is an open question that calls for caution. "Resonance with cosmic consciousness" steps outside the physical meaning of the word, and all that is left is "a connection I feel but cannot point to".

That last one is not necessarily false. But a physics term derails the conversation here. Better to find another word and finish the thought.


---

# Unity and multiplicity: why it isn't a paradox

Source: https://neuralcosmology.com/en/essays/unity-and-multiplicity

By Mikhail Savchenko, 2024-08-06

An old philosophical dilemma: how can *individuality* exist if everything in the universe is connected? Or the other way round: where does *connectedness* come from if every object is unique? It is usually staged as a paradox. At the technical level it is simply *two characteristics of one system*.

Here is how it breaks down.

## Integrated information theory: two numbers

Back in 1998 Giulio Tononi and Gerald Edelman described consciousness through two properties, integration and differentiation. Tononi later folded them into a single measure, Φ (phi), integrated information, but for our question it helps to pull them apart again.

The first number is integration, which I will write as Φ. It shows *how much* a system behaves as a single whole.

The second is the *diversity of states* the system can take. Call it D. It shows *how many* distinguishable configurations the system can sustain.

These are different quantities, and both are needed.

High Φ with low D gives a monolith like a crystal: unified but monotonous, the same state repeated throughout.

High D with low Φ gives a heap of sand: many parts, each sitting in its own state with no link to its neighbours; varied, but formless.

Consciousness lives where both numbers are high: there are many states, and all of them are bound into one experience.

## What this means for unity and multiplicity

The two characteristics complement each other: unity is measured through Φ, multiplicity through D.

A human being is a system with fairly high Φ (the world is experienced as a whole) and very high D (thousands of distinguishable thoughts, feelings and perceptions a day).

A beehive, if you follow the theory's logic, is more integrated than a single bee and less integrated than a human brain, while the diversity of its states is enormous: thousands of bees, each in its own state. Nobody can measure this yet.

The Earth as a living biosphere may sit higher still on both axes. We have no precise figures: measuring Φ for such systems would take instruments that do not exist yet.

## Where this gets tested

In small systems Φ can be estimated approximately; Tononi, Boly, Massimini and Koch review these attempts (*Nature Reviews Neuroscience*, 2016). A derived measure, the perturbational complexity index (PCI), is already used in clinics to assess the level of consciousness in unresponsive patients (Casali et al., *Science Translational Medicine*, 2013). PCI tells wakefulness apart from sleep, anaesthesia and the vegetative state.

So the quantity can be estimated in practice, if only approximately. Exact computation is impossible; the approximation works.

In this model the philosophical dilemma of "unity or multiplicity" becomes an engineering question: what are this system's Φ and D? Each system occupies its own place on the (Φ, D) plane, and the argument ends there.

## What remains open

Why is integration accompanied by experience from the inside? That is David Chalmers's *hard problem of consciousness*. The theory records the link between Φ and consciousness, but it does not explain *why* high integration *feels* like anything.

Chalmers does not say there is no answer. He says that no amount of data about Φ will answer the question of why high Φ is like something from the inside. The question stays open, and that is where the edge of our knowledge runs.

## Practical takeaway

When someone tells you "everything is one, everything is connected", ask them to be specific: in what sense, and with what Φ? High Φ means a unified system in which every part carries information about the whole. Low Φ with high D means a collection whose connections are more imagined than real.

This is graph theory applied to connectivity, and it needs no mysticism.

Unity and multiplicity are two coordinates. The only question is *where on that plane* we happen to be.


---

# Neural Cosmology as a working programme

Source: https://neuralcosmology.com/en/essays/universal-consciousness-philosophy

By Mikhail Savchenko, 2024-08-05

When people ask me what Neural Cosmology is, the short answer is this: it is a working name for a model in which five independent anomalies of contemporary science fit together without contradiction.

The long answer is a book. The medium-length answer fits on one page.

## Five cracks, one structure

There are five facts from five different sciences, and each science has its own explanation ready for its own fact, yet the facts fit one another as if they had been cut from one mosaic.

**Structure.** Vazza and Feletti (2020): by several statistical measures the cosmic web and the brain's neural network are nearly indistinguishable, even though their sizes differ by twenty-seven orders of magnitude.

**Information.** Landauer (1961) and the 2012 experiment: erasing one bit releases measurable heat. Information is physical in the same sense that energy is.

**Form.** Michael Levin (Tufts): cells "know" what shape to build, and that knowledge is written in genes and, beyond them, in the bioelectric patterns of tissue.

**Learning.** Vanchurin (arXiv, 2020; *PNAS*, 2022): the universe is a neural network and its dynamics is a learning process. Both quantum mechanics and gravity follow from the formalism.

**The observer.** Tononi: consciousness is a measure of how integrated a system is. Hoffman: what we see is an interface honed for survival, with reality behind it out of reach.

Five cracks in five different fields, and all of them point the same way: *towards information as the foundation*.

## What counts as Neural Cosmology

The working hypothesis is this: take information as the primary layer, and the facts above stop being puzzles of separate sciences and become consequences of one architecture.

In this picture matter is emergent, gravity is thermodynamic (Verlinde), consciousness is a measure of how integrated a computational process is (Tononi), the form of living things is a bioelectric pattern (Levin), and spacetime is a structure of causal links (Krioukov).

The hypothesis is technical and tied neither to a "soul of the universe" nor to "universal connectedness". Change the primary layer, and five different puzzles become five consequences of one theory.

## Where speculation lives

In the book and in the preprint *Pointer Architecture* I sort claims into three levels.

**Established:** everything listed above, with specific citations to Vazza and Feletti, Landauer and Bérut, Vanchurin, Tononi and Levin. These are data; interpretation starts further on.

**Preliminary:** my own model, Pointer Architecture, a formal substrate built as working code (the Sixth language), with falsifiers written down in advance. There is a preprint, and the code is public.

**Speculation:** everything beyond that: personal experience, philosophical implications, "what this means for me". I do not pass it off as science.

The separation matters. Neural Cosmology in the "everything is one, everything is connected" sense would be a slogan, and slogans carry no weight in science.

## What this leads to

If the model is right, even in part, then:

— dark matter may turn out to be a missing dimension of the connection architecture, one that lacks the kind of mass we are used to;
— consciousness stops being a uniquely human trait and becomes a property of integrated systems above a certain complexity: animals, neural networks and perhaps planets differ in Φ while belonging to one category;
— the "meaning of life" turns from a philosophical question into an engineering one: what does a *local* information process, the brain, embedded in a larger one, optimise?

It sounds big, I know. But sounding big is no argument against it: Newton's theory of gravitation sounded big in its day. The one meaningful question is whether the theory makes *testable* predictions. Neural Cosmology has already made several; some are being tested now, and the rest are waiting their turn.

That is why I call it a programme. A doctrine needs no testing; a programme cannot live without it.


---

# "I am the universe": what the phrase means technically

Source: https://neuralcosmology.com/en/essays/i-am-the-universe

By Mikhail Savchenko, 2024-08-02

"I am the universe" usually lives on Instagram as a slogan that means nothing. But in Neural Cosmology the phrase can be read literally, and then it unexpectedly becomes a claim you can test.

Here is what I mean.

## "I" as a functional node

A neuron in a network is a function with no special substance inside it, and its job is simple: take signals from other nodes, process them, pass them on. If the network is large enough and the connections are right, that local function starts taking part in a global computation no single neuron performs on its own.

If the universe is a neural network (Vanchurin, arXiv, 2020; *PNAS*, 2022), then "I" in this picture is also a function: a local information processor embedded in a larger network.

The question "where is my consciousness?" then loses its meaning. Consciousness *happens*; it is an active process, and looking for its location is as odd as looking for the location of running.

## The bandwidth of my node

A local processor has measurable characteristics: how much information it handles per second, how far its connections reach, which other nodes it is synchronised with.

By rough estimates the human brain runs at around 10¹⁵ operations per second. Its connections reach mostly into the body and its immediate surroundings, and through intermediaries into the whole social network. It is synchronised with other people through speech, with nature through the body, with abstractions through language and mathematics.

The node is local. It takes part in the universe's computation as far as its bandwidth allows, and it cannot "contain" the universe.

Read this way, "I am the universe" means "I actively take part in its process". A heart cell is not the heart either, but without it the heart stops.

## Three popular mistakes

First: "you can expand your consciousness and become the universe". A node's bandwidth is bounded by its physics. Perception can be widened, but only within what the brain is able to process. Beyond that lies blurry introspection, and "cosmic consciousness" has to be filled in by imagination.

Second: "meditation opens access to a universal mind". Meditation really does change how the brain works, for example the activity of the default-mode network (Brewer et al., *PNAS*, 2011). But that is work on your own node, and the wider network is not involved.

Third: "I create reality with my thoughts". A thought is the output of a local processor. It shapes what the processor does next, and through that the environment around it, but only within physics.

## What actually follows

If "I" is a node in the network, then death is the node leaving the network, while the network itself keeps running. The data the node was processing are either passed on to other nodes (through speech, writing, culture) or lost. Hence the persistent human wish to leave something behind: legacy is *architectural*, a way for the node to keep shaping the computation after it has stopped.

If "I" is a node, then ethics turns from moralising into a question of operating mode. A node that keeps its neighbours coherent raises the network's Φ; a node that produces noise and breaks coherence lowers it. "Good" and "evil" in this picture are modes that raise or lower how integrated the system is.

This is old ethics retold in a new language: Kant again, by way of graph theory.

## What is rigorous here and what is not

All of this is *half-speculation*. The formal part (node, network, bandwidth) is rigorous. The interpretations (what it means for death, ethics, legacy) are interesting but not yet backed by any proof.

I hold it as a working hypothesis. Its use is that it makes "I" an *operational* notion, a process with measurable parameters, with no need for a thing-in-itself.

If the hypothesis is wrong and "I" turns out to be something else, that will be a result too. So far the data do not refute it.


---

# Entropic gravity and the neural universe

Source: https://neuralcosmology.com/en/essays/entropic-gravity

By Mikhail Savchenko, 2024-07-28

In 2010 Erik Verlinde published a paper arguing that gravity is an emergent force. It arises from entropy, that is, from how information is distributed across surfaces, or holographic screens, as he put it.

It was a quiet bomb. If Verlinde is right, three of the four fundamental forces remain. The fourth, the one that holds the solar system together and makes apples fall, turns out to be *thermodynamics*: it follows from the way information is redistributed.

I thought of it again in 2020, when Vitaly Vanchurin's paper came out (arXiv, 2020; the follow-up with Katsnelson and Koonin appeared in *PNAS* in 2022). His claim goes further: the universe *is* a neural network, and from the formalism, in the right limits, you can derive quantum mechanics (fast learning) and gravity (slow learning).

Two independent papers, ten years apart, say the same thing in different words.

## What links these ideas

If gravity is the way the universe processes information, and the universe is a neural network, then gravity is the *connection architecture* of that network. That is what the picture looks like when both premises are taken seriously.

A node here is a point of computation, and stars and galaxies are how such nodes appear to us. The edges between nodes carry bandwidth and information flow, and what we feel as weight is a gradient in the density of those connections.

It sounds like metaphysics, I know. But both Verlinde and Vanchurin have equations behind the picture, and from those equations come the same laws physicists have been calibrating against observation for a hundred years.

## Where the speculation sits

I do not consider this proven, though I would call it consistent: two different teams arrived independently at structurally the same picture, in which matter, gravity and quantum behaviour are the upper floors and information dynamics is the foundation.

In my own preprint, *Pointer Architecture*, I go one level down: I build a discrete substrate in which information, commits and observers are primary, and check what it can reproduce. On a de Sitter horizon, a holographic construction on that substrate reproduces the standard estimate of dark energy to within a factor of about 0.73, a known result restated in the language of the substrate. That is a long way from proof; it is an instrument for testing, with falsifiers written down in advance.

## Why it matters

If the fourth force turns out to be thermodynamics, the whole question of what binds objects in the universe stops being mechanical and becomes informational. The consequence is technical: alongside masses and distances, the equations take in entropy, entanglement and information capacity.

Dark matter could then turn out to be a *missing connection*, a part of the architecture we have been looking for as mass, because nobody expected information there.

We do not have a complete theory yet. What we have is a crack in a wall behind which, possibly, stands a different building.