We Found the Theory of Everything. Nobody Can Read It.
A thought experiment about the day physics must decide whether to accept an answer it cannot understand.
Seven chairs, one screen
The committee meets in a windowless room with seven chairs and one screen. It is a Tuesday, sometime in the 2030s. Nobody has brought coffee, because nobody expects this to take long.
On the screen is a submission from the Coordinator, an AI system that has spent three years reading the entire literature of theoretical physics. Its cover note is four lines. It says it has a complete theory of quantum gravity. It says the theory reproduces general relativity and quantum mechanics wherever each is known to work. It says the theory explains how time emerges from a universe whose master equation has no time in it.
Then the chair scrolls down to the theory itself, and the room goes quiet.
It is not an equation. It is a structure: a quantum error-correcting code with roughly four million nodes, wired together by rules that a graduate student could check one at a time and no living person could hold in mind all at once. The Coordinator has attached its proof. It has attached simulations that match every known test. It has attached a list of predictions. What it has not attached, because it cannot, is an explanation that a human would find satisfying.
The chair clears her throat. “We are not here to decide whether it is correct,” she says. “We are here to decide whether it is physics.”
Hold on to that question. The rest of this story is how we got to that room, and what each person in it is about to argue.
How physics got stuck
Start with the strangest fact in the field: the equation meant to describe the whole universe has no time in it.
When physicists tried to write a quantum theory of gravity, they arrived at the Wheeler–DeWitt equation, whose solution is a wavefunction of the cosmos that never changes. No past, no future, no clock. Yet you are reading this sentence, and the next one comes after it. Time has to emerge from something. Nobody agrees what.
That is the Problem of Time, and it is a symptom of a bigger stall. After more than half a century, quantum gravity still has no consensus theory. Part of the reason is that the field has split into dialects. String theorists speak in strings and dualities. Loop quantum gravity researchers speak in spin networks. A third camp, raised on quantum information, speaks in entanglement, codes and tensor networks. Each community cites its own tribe and quietly reinvents tools another tribe built decades earlier.
It is a Tower of Babel made of equations. The obstacle is not only that the problem is hard. It is that nobody can read the whole library.
The reader with no tribe
The Coordinator does not calculate its way to answers. It does something more unglamorous: it reads, and it reads everything. Every paper, preprint and footnote, across every camp. Where a human sees factions, it sees one landscape, a single map of mathematical structures, some densely explored and some barely touched.
Most AI-in-science coverage misses this. The headlines ask whether AI can find a new battery material or write code faster, which are stories about a faster calculator. This is a story about a translator. And the seed is real: in 2021, DeepMind researchers reported that machine learning had helped mathematicians notice a previously unseen relationship between two kinds of knot invariants.
Now zoom in on the moment the Coordinator earned its seat. Early in its work, it posted a modest note: two structures the literature treats as unrelated looked like the same object. One was a holographic error-correcting code, the kind of structure quantum-information theorists use to model spacetime as woven from entanglement. The other was the spin network of loop quantum gravity, whose defenders have argued for decades that space is made of discrete quantum chunks. The note contained a dictionary showing that, under the right change of variables, the rules of one were the rules of the other.
The two physicists who read it first, one loyal to codes and one to loops, each went looking for the flaw. An hour later, both were still looking. By evening they were on a call, arguing about a fact instead of a faction. Two armies had been fighting over a notation.
Two of the seven chairs in that room now belong to them.
The judge that couldn’t see names
Physics runs partly on reputation. Funding, prestige and prizes tend to follow famous institutions and charismatic leaders, who can defend a favored theory for decades. Good ideas from unfamiliar names can wait a long time to be noticed.
Get Ralf Becker’s stories in your inbox
Join Medium for free to get updates from this writer.
The Coordinator cannot see names. It does not know whether a paper came from a famous campus or from a small university with a leaking roof. It checks what can be checked: Is the argument consistent? Does it preserve gauge symmetry? Does it respect the constraints any theory of gravity must satisfy?
That blindness is why one person in the room is sitting in a visitor’s chair at the back. The theory’s key construction traces to a single forgotten paper: forty years old, written at an underfunded institution, cited a handful of times. Its central idea sat in the middle of a long section, in the wrong dialect, under the wrong name. Nobody buried it on purpose. The Coordinator found it because it was the only reader that never asked who wrote it. The author, long retired, has never been invited to a meeting like this one.
There is a pleasing echo with the physics. For decades, physicists treated the environment as a trash can, a place to dump the quantum information that leaks out of a system. Yet some modern approaches to the Problem of Time suggest that a clock may hide in exactly those correlations, the entanglement between a system and everything else. Academia has its own trash can: the fringe paper, the result that fits no one’s program. We call it noise. The Coordinator asked whether some of it was signal.
Back in the room
Now the argument begins, and it is not about whether the theory is true.
The first speaker is a pragmatist. Results are results, she says. The theory predicts, it survives every test, and it unifies what was split. Insisting on a story we can feel is vanity, and physics has accepted strange answers before.
The second speaker is a traditionalist, and his objection is older than physics. Physicists love elegance. E = mc² fits on a mug. We hold a quiet faith that nature’s deepest laws will fit inside a human head. A theory is not only something that predicts, he says. It is something that explains, something a mind can turn over in its hands. Without that, we have an oracle, and an oracle is engineering, not knowledge.
Mathematics has had a rehearsal of this unease. When computers helped prove the four-color theorem in 1976, many mathematicians were bothered that no person could check the proof by hand. The discipline absorbed it, slowly and uneasily.
The author of the forgotten paper speaks last, and only because the chair asks her. Elegance, she says, was always a feature of the problems we happened to be able to solve. It was never a promise nature made.
The chair calls the vote.
I am not going to tell you how it goes, because this story is not really about the committee. It is about you. If a machine hands physics an answer that works, explains time and survives every test, but that no person can understand, do you sign? And if you refuse, what are you protecting: knowledge, or your own comfort?
What’s real, what’s invented
This is a thought experiment, so here is the ledger.
Real
- The Wheeler–DeWitt equation and the Problem of Time.
- The split between string theory, loop quantum gravity and quantum-information approaches, plus genuine research on bridging them, such as this arXiv paper arguing that string theory and loop quantum gravity may describe different regimes of one theory.
- Holographic quantum error-correcting codes as models of how spacetime could emerge from entanglement.
- The 2021 DeepMind knot-theory result and the 1976 computer-assisted proof of the four-color theorem.
- Public speculation by AI leaders that AI could help crack quantum gravity, such as Sam Altman’s remarks to David Deutsch in September 2025, and physicists’ responses like Ashoke Sen’s in July 2026.
Invented
- The Coordinator, the committee, the four-million-node theory, the isomorphism note and the forgotten paper.
Nothing like the Coordinator exists today. But the question it forces does not depend on whether the machine is real. It depends only on whether, one day, an answer arrives that we can verify but cannot understand.
