Physics has a number that Richard Feynman called "magic," and nobody knows why it is what it is. This week a physicist published a formula that hits it to ten digits.
I'm not a physics expert. But the claim is easy to state, and it comes with a proof file anyone can run. I wanted to know how much of it holds up. So I read the paper, and an AI assistant helped me check the numbers and draft this post. Here's what we found.
The number everyone puts on the wall
The number is called the fine-structure constant. Its symbol is α (alpha), and it's close to 1/137. It sets how strongly light and electric charge push on each other.
It has no units, so it's the same whatever ruler you use. And we know it very well, to about one part in ten billion.
What we don't have is a reason for it. Today's best theory of particles, the Standard Model, takes this value from experiment. It has about two dozen numbers like that, and nothing inside the theory explains their values.
Alex Wissner-Gross opens that part of his post with a Feynman line. Feynman wrote that "all good theoretical physicists put this number up on their wall and worry about it." He called it "a magic number that comes to us with no understanding by man."
That's why a derivation would be a big deal. If a theory produced α from scratch, with nothing fitted, it would have to contain something new that the Standard Model lacks.
What was claimed
On October 24, 2026, Wissner-Gross announced what he calls Chain Theory. The preprint is dated October 23. It starts from one sentence, 18 words long:
"Nothing exists but finitely many maximally entangled pairs of finite-dimensional systems, some shared, known only by distilling them."
In plain words: the world is made of pairs of tiny quantum systems that are tied together by entanglement. Entanglement means two things share a state, so you can't fully describe one without the other. He pictures particles as chainmail, a mesh of such pairs as links.
The smallest link that can be "read" joins a three-level system to a two-level one. He says the symmetries of that pair have the same shape as the Standard Model's force groups. That part caught my eye.
Then comes the number. His readout counts 172 "measurement records" of that one pair. They sort into four piles of 79, 48, 42 and 3 records. Each pile gets a weight: 1, √2, √3 and √6. Add it up, divide by 31104, and you get:
(79 + 48√2 + 42√3 + 3√6) / 31104
Flip it upside down and it's 137.035999189566… The official CODATA 2022 value, which is a world average of the best measurements, is 137.035999177, give or take 0.000000021. Wissner-Gross says his number lands 0.6 standard deviations away.
A standard deviation, or sigma (σ), is the size of the measurement's error bar. Being 1σ off is routine. Physicists want about 5σ before they call something a real difference. So 0.6σ is a very good match.
I'll be honest. I can't judge the physics. I can only look at what's checkable, and at what the author says himself.
What we could check
The assistant rebuilt the number from the paper's own definitions. The value matches the paper to every digit it prints. The count of 172 records, split 79, 48, 42 and 3, also comes out right. And the 0.6σ gap to CODATA is correct.
The paper also ships a file for Lean. Lean is a program that checks proofs line by line, with no room for hand-waving. The assistant ran it. It built without errors, with no unfinished steps ("sorry") and no extra assumptions beyond Lean's standard ones.
So the bookkeeping is sound. That's worth saying. Many number-from-nowhere claims can't clear even that bar.
But look at what Lean checked. It checked that the author's definitions give that formula. It did not check that the definitions describe nature. The paper says as much: the file doesn't cover the link to the real α, which the author calls the Photon Conjecture. That link is the one step that is only a guess.
Why I'm cautious
Here's where I slow down. None of this is a knock on the author. He flags most of it himself.
The biggest gap is the leap itself. The theory says this qutrit-qubit pair is the link photons are made of, so its number is α. That is a proposal, not something derived. The paper says its model builds no propagator and no function of momentum. It calls the match "a correspondence of roles, not of diagrams."
Next, the measurements don't agree with each other. Look at the chart below. The electron magnetic moment gives one value, rubidium atoms give another, and cesium atoms a third. The formula sits +1.6σ, −1.5σ and +5.3σ from them. It's close to the average, and far from the cesium result. The 0.6σ is a match to a blend of numbers that argue with each other.
Values from the paper's Fig. 3 and text. Chart is mine, redrawn from those numbers.
Then there's a trap that catches careful people too. It's called the look-elsewhere effect. Think of finding your birthday in the digits of pi. Search long enough and you will. The assistant estimated how many forms like this one, whole numbers with √2, √3 and √6 over a denominator up to 31104, land that close to the measured value. The answer was about 140. It's a rough count, but it means a ten-digit match isn't shocking by itself.
To be fair to the author, he didn't go hunting. The number comes out of a fixed set of steps. The worry is about choices inside those steps, like the weights and dividing by 31104 instead of 172. The paper lists other ways of wording those choices. They give very different values, and two of them land within one percent of the answer. It says each is ruled out by a definition, not by comparing with α. I can't tell the order he made those choices in, and neither can anyone else from outside.
And there's only one number. Beyond α, the assistant found no second figure in the paper that you could check against a measurement. The assistant only skimmed the very long working draft, but it says no reading yet lands on the measured proton-to-electron mass ratio. One hit is exciting. A second would be much more convincing.
Last, it's new and self-published. There's been no peer review yet. The paper also says plainly that the work had help from "resources of Physical Superintelligence PBC, in which the author has a financial interest." I'm glad he disclosed it. Readers should know.
In Wissner-Gross's own words: "I do not yet claim to have done the same," about rebuilding physics, and "much of it is still conditional or underdeveloped." I respect that.
How this could get settled
Here is the good part. The number is exact. It has no knobs left to turn. So new measurements can test it.
The post names three: the electron's magnetic moment at Northwestern University, and atom-recoil experiments at Berkeley and in Paris. They aim to be up to ten times more precise. If they agree with each other, they will either close in on his number or rule it out. Taken alone, the cesium result already rules the formula out at 5.3σ, so which side is right matters a lot.
That's what I like most about this claim. It can lose.
My take
My view is that this is interesting but unverified. The math is real, and anyone can check it. Whether it's physics is a different question, and only independent physicists and better measurements can answer that.
I'm sharing this because a bold claim like this deserves a careful look in plain words. And because I'm sure some Hive readers know this field far better than I do.
If you know this field, tell me in the comments where I got something wrong, or what you'd check next. If you show me an error, I'll correct the post.
Sources: Wissner-Gross's post · Preprint (PDF) · Lean file · Working draft (PDF) · CODATA 2022 value of α
AI disclosure: an AI assistant helped research and check the numbers, helped write this text, and helped make the cover and the chart. I reviewed it before posting.