In this article, I revisit my tongue-in-cheek attempt to use AI to explore grand unification - but this time with a narrower, more testable idea. Instead of trying to solve all of physics in one heroic act of optimistic nonsense, I look at whether dark matter and dark energy could be two linked behaviours of spacetime's hidden quantum-information structure. It is speculative, probably flawed, and not peer-reviewed physics, but that is partly the point: this is an experiment in how far AI can help a non-specialist reason into difficult scientific territory.
Skip the reflection and read the follow-up paper now.
In September 2024, shortly after OpenAI released its first o1 reasoning model, I did what any sensible non-physicist would do: I asked it to help me solve one of the hardest problems in science.
Naturally.
I wrote about that experiment in my previous article, How I Unified General Relativity with Quantum Mechanics (Grand Unification With AI). The title was deliberately tongue-in-cheek, because no, I did not actually solve quantum gravity from my laptop after watching a suspicious amount of PBS Space Time. I was rooting for me too, but the universe remains stubborn.
What I did do was use a frontier AI reasoning model to explore a big theoretical question: could spacetime, dark matter, and dark energy all be connected through quantum entanglement?
The first version of the paper was ambitious. Very ambitious. It tried to connect emergent spacetime, de Sitter holography, dark matter, dark energy, gravitational waves, higher-spin fields, and the Standard Model into one large structure. In hindsight, it was probably trying to carry too much weight. That is a polite way of saying it was attempting to solve too many hard problems at once, which is the classic beginner move. Why fail at one impossible task when you can fail at seven?
Since then, I have revisited the idea with newer AI models, more critique, and a much stronger appreciation for how easy it is to produce something that sounds impressive while still being wrong in subtle ways.
The result is a new, narrower paper:
Entanglement Disequilibrium as a Unified Dark Sector: A Relative-Entropy Response EFT for Dark Energy and Axion-like Dark Matter
(This time generated by iterating with OpenAI's GPT-5.5 Pro reasoning model)
The honest disclaimer
I should say the important bit upfront.
I am not a peer-reviewed physicist. I am not a mathematician. My formal background is computer science and business management, and my day job is running a software development company, not deriving cosmological perturbation equations on a whiteboard while looking haunted.
So this article should not be read as:
A software company director has solved dark matter and dark energy.
It should be read as:
A technically-minded non-specialist used advanced AI models to explore a speculative theoretical physics idea, narrowed it into a more disciplined mathematical proposal, and is publishing the result as an experiment in AI-assisted reasoning.
That distinction matters.
The paper may contain mistakes. It may contain glaring mistakes. It may contain the sort of mistake that makes an actual cosmologist take off their glasses, rub the bridge of their nose, and quietly reconsider the future of civilisation.
That is entirely possible.
But I still think the experiment is worth documenting, because it shows something important about where AI is heading. These models are becoming capable of helping non-experts enter difficult intellectual territory, structure ideas, test assumptions, find related work, and produce something clear enough to be criticised properly.
That last part matters. The goal is not to create something beyond criticism. The goal is to create something precise enough that criticism can land.
From "grand unification" to a narrower question
The first paper tried to do too much.
It started from a serious idea in modern theoretical physics: perhaps spacetime is not fundamental. Perhaps the geometry of spacetime emerges from quantum entanglement.
That is not my idea. There is a real body of physics exploring connections between entanglement, geometry, gravity, black holes, holography, relative entropy, and quantum information. My earlier experiment borrowed from that broad area and then attempted to push it into an enormous unifying framework.
The new paper is more restrained.
It asks a narrower question:
If ordinary gravity is what spacetime looks like when quantum entanglement is in equilibrium, what might we observe when that entanglement structure is slightly out of equilibrium on cosmic scales?
Rather than trying to solve quantum gravity, the new paper focuses on the dark sector: dark matter and dark energy.
The simple version of the idea
The universe appears to contain two huge mysteries.
First, there is dark matter. Galaxies rotate and cluster as though there is more gravitating matter than we can see. Something invisible appears to help hold cosmic structure together.
Second, there is dark energy. The expansion of the universe appears to be accelerating, as though some smooth energy component is pushing the universe apart on large scales.
In standard cosmology, these are usually treated as separate things.
Dark matter is one mystery. Dark energy is another.
The new paper explores a different possibility:
What if dark matter and dark energy are two different behaviours of the same underlying spacetime-information response?
The easiest way to picture it is as a cosmic dial.
The distance from the centre of the dial behaves like dark energy.
The angle around the dial behaves like dark matter.
In the paper, this is written mathematically as:
Phi_E = rho e^(i theta)
The radial part, rho, behaves like dark energy when it changes slowly.
The phase part, theta, behaves like axion-like dark matter once it starts oscillating.
In plain English:
Dark energy may be the slow "stretch" of an entanglement-response field, while dark matter may be the "rotation" or phase of that same field.
That is the central idea.
What might be novel?
This is where it is important not to overclaim.
The individual ingredients are not new. Physicists have already studied gravity emerging from entanglement, relative entropy in gravitational systems, axion-like dark matter, quintessence, interacting dark energy, and complex scalar fields in cosmology.
So the paper is not saying:
Nobody has ever thought about any of this before.
They absolutely have.
The potentially novel part is the way these ideas are connected.
The paper proposes this chain:
relative-entropy residual -> positive response metric -> polar entanglement-response field -> linked dark matter and dark energy behaviour
In normal language:
If spacetime's quantum information structure has a measurable "out of balance" response, then the simplest two-part version of that response may naturally split into one part that looks like dark energy and another part that looks like dark matter.
That is the distinctive move.
The paper does not claim that complex scalar fields are new. It explicitly says they are not. The claim is narrower: perhaps a relative-entropy or entanglement-disequilibrium argument can motivate a polar response field where dark matter and dark energy are connected rather than independent.
Why relative entropy?
Relative entropy is a mathematical way of measuring how distinguishable one state is from another. In quantum theory, it is always positive, which makes it useful as a measure of departure from a reference state.
The paper starts from the idea that local spacetime geometry may correspond to an equilibrium condition of quantum entanglement.
At equilibrium, you get something like ordinary Einstein gravity.
Away from equilibrium, there may be a residual: a positive "leftover" quantity that measures how much the local entanglement structure differs from the equilibrium state.
The paper then treats this residual as an effective response system at very large scales.
That phrase, "effective response system", is doing a lot of work. It means the paper is not claiming to know the microscopic quantum-gravity theory underneath everything. Instead, it asks what kind of large-scale field theory could describe the behaviour if such a residual exists.
That is similar in spirit to studying waves in water without modelling every molecule of water. You can describe the wave before you know every microscopic detail.
Of course, the difficult question is whether this particular response system is actually justified. That is where the model is most vulnerable, and where real physicists would need to attack, test, or improve it.
The key point: the model gives itself ways to fail
The most useful thing about the new paper is that it does not merely say:
Maybe dark matter and dark energy are connected somehow.
That would be too vague to do much work.
Instead, in one branch of the model, the connection is mathematically fixed.
If the dark-energy part changes over time, then the dark-matter part should also change in a related way. The paper describes this as a late-time consistency triangle:
dark energy evolution <-> dark-sector coupling <-> dark matter mass drift
In more human language:
If dark energy is rolling or changing, then dark matter should dilute slightly differently, and the effective mass of the dark matter field should drift in a predictable way.
That matters because science needs failure modes.
If future cosmological observations find that dark energy evolves in one way, but dark matter structure does not show the linked behaviour predicted by the model, then this version of the theory fails.
That is not a weakness. That is the point. A speculative theory that cannot be killed is usually not doing much scientific work.
A possible link to primordial gravitational waves
The paper also explores a connection with inflation, the extremely early period when the universe is thought to have expanded incredibly rapidly.
In one branch of the model, if the phase field existed during inflation and makes all of the dark matter, then the model predicts a very small upper limit on primordial gravitational waves.
Put simply:
If this version of the model is true, future experiments probably should not see a strong primordial gravitational-wave signal.
More specifically, a robust future detection at around r ~ 10^-3 would rule out that particular branch.
That would not automatically kill the whole idea, because other branches could avoid that constraint. But it would kill a specific version of it, which is exactly what a useful model should allow.
Why this is not just AI hallucination with equations
This is the awkward bit.
Large language models are very good at producing convincing text. That is both powerful and dangerous.
In a field like theoretical physics, a model can generate equations, cite concepts, and sound confident while being wrong. That is why this experiment is interesting and risky.
The process was not simply:
Write me a theory.
It was iterative. I asked the model to generate possible formulations, criticise its own assumptions, compare the proposal to existing physics, identify prior art, find failure modes, narrow overambitious claims, derive consequences, and explain what would falsify the theory.
That does not guarantee correctness. It does, however, move the output away from vague speculation and towards something more structured.
The biggest improvement from the first paper to this one is not that the maths became more impressive. It is that the claim became smaller, clearer, and easier to attack.
The first version tried to be a grand theory.
This version says:
Here is a speculative effective theory for the dark sector. Here are the assumptions. Here is what is not new. Here is what may be new. Here is how the model could fail.
That is a much healthier shape for an idea.
The Dunning-Kruger problem gets more serious with AI
One of the themes in my original article was the Dunning-Kruger effect: people with limited expertise often struggle to assess their own competence.
AI makes this problem more complicated.
When an AI model gives you a fluent, confident, equation-filled explanation, it is easy to feel as though you understand more than you do. The output can look expert-level even when the underlying reasoning contains hidden mistakes.
That is why I think the right posture is:
- curiosity without arrogance;
- ambition without pretending certainty;
- clear separation between "this is interesting" and "this is true";
- openness to expert criticism.
I am not publishing this because I think I have personally solved the dark universe.
I am publishing it because the process itself is a fascinating example of where AI-assisted exploration is heading.
What I think is genuinely interesting
The most interesting part is not whether this exact paper turns out to be right.
Realistically, it probably has problems. That is what happens when a non-specialist uses AI to wander into theoretical cosmology wearing hiking boots and an optimistic expression.
The interesting part is that AI made it possible to explore the idea at a level of structure and mathematical detail that would previously have been completely inaccessible to me.
I could start with an intuitive question:
Could dark matter and dark energy be connected to entanglement?
Then I could keep pushing:
- What would the field be?
- What equations follow?
- What does this resemble in existing physics?
- What is not original?
- What could be tested?
- What would kill it?
- What are the obvious objections?
- How do we avoid overstating the claim?
That is not the same as becoming an expert. It is not a replacement for peer review. But it is a new kind of intellectual tool.
It lets a curious person climb further up the mountain than they otherwise could, even if they still need an experienced mountaineer to point out that they are standing on a cornice.
What would it mean if the paper were right?
If the model turned out to be pointing in the right direction, the implication would be profound.
It would suggest that dark matter and dark energy are not separate cosmic ingredients, but two linked behaviours of the same underlying quantum-information structure.
It would mean that ordinary gravity is something like the equilibrium behaviour of spacetime, while the dark sector is what we observe when that equilibrium is slightly disturbed.
That would make the dark sector less like "unknown stuff added to the universe" and more like "the visible response of spacetime's hidden information architecture".
That is a beautiful idea.
It may be wrong. But it is beautiful enough, and structured enough, to be worth exploring.
What would need to happen next?
For the paper to become genuinely serious, two things would need to happen.
First, a physicist would need to examine the theoretical foundations, especially the assumption that the relative-entropy response should contain a compact phase coordinate.
Second, the model would need to be implemented in cosmological tools such as CLASS or CAMB, so it could be tested properly against observations: the cosmic microwave background, galaxy clustering, weak lensing, supernovae, baryon acoustic oscillations, and structure-formation data.
That is where the model either becomes more interesting or gets crushed by reality.
Reality, historically, has a fairly strong record.
Final thoughts
This project started as a playful experiment: could I use AI to explore one of the hardest problems in physics from the outside?
The first attempt was wildly ambitious and deliberately framed with a wink.
This second attempt is more restrained. It does not claim to solve quantum gravity. It does not claim to prove what dark matter and dark energy are. It proposes a specific, speculative, testable way they might be connected.
That feels like progress.
Not necessarily progress towards being right.
But progress towards asking a better question.
And that, to me, is the real value of the experiment.
