Could You Compare Both Sides of a Horizon?
The Contradiction Is Real, and Nobody Will Ever Hold It
Drop a detector into a black hole and keep another one outside. Record what the outside one sees and keep the record. Wait for the hole to evaporate, however long that takes, and reconstruct the inside detector’s record from the radiation. Put the two records side by side. Everyone says they cannot both be true. Show me the contradiction.
The request is taken exactly as stated and physics answers it one wall at a time, each wall marked by how firmly it stands. This page leans on three others: Black Holes for what a horizon is, Could You Hold a Black Hole for why nothing crosses back, and The Top of Reality for the number of operations universe can ever perform. That last number turns out to be the one that matters. After the walls comes something new for this section: the objections a sharp reader raised to a draft of the walls, and what physics says back.
What Physics Says, Wall by Wall
Pick one photon of the hole’s glow, emitted late, after more than half the hole has evaporated. Call it B. The theory says two things about it.
First: the horizon is smooth, and a detector crossing it notices nothing, which is the equivalence principle at work. For that to be true, B must be born as one half of a perfectly entangled pair with a partner A just inside the horizon, the way two photons from one source are. A perfectly entangled pair is sealed: if B and A together are in a pure state, B carries no correlation with anything else in universe. That is a theorem, not an assumption.
Second: information is not destroyed. Then the radiation as a whole must end in a pure state, which forces the late radiation to be entangled with the early radiation, R. Don Page proved in 1993 that after the halfway point every late photon has to be nearly perfectly entangled with what came out before. So B must be sealed with R.
One photon cannot be sealed with two different partners. That is not difficult, it is impossible, in the way a number cannot equal two different numbers. So before any detector is dropped, three ordinary assumptions, a smooth horizon, no loss of information, and ordinary field theory near the horizon, contradict one another. The request did not create the contradiction. It found the place where it is already sitting.
Drop the smooth horizon and you get a firewall: a detector crossing burns. The equivalence principle, the best-tested idea in gravity, then fails at a surface whose position is defined by the entire future of universe rather than by anything local. Drop the conservation of information and quantum mechanics, the most precisely tested theory there is, becomes an approximation. Drop the independence of inside and outside, and the partner A inside the horizon is not a separate system at all but the early radiation R, described a second time. Then B is sealed with one thing, not two, and there is no contradiction. The price is that inside is no longer a place of its own.
The third exit was words until 2019. Then several groups showed that the entropy of the radiation, computed with the interior counted as part of it in the right way, rises and falls exactly as conservation of information requires. The calculation is in simplified models, mostly two-dimensional, and far from a real star. But it was the first time the answer came out of an equation rather than a preference, and it is why the third exit is now the favourite.
Here is why the contradiction has never been seen and, as far as anyone can tell, never will be. To show that B is sealed with the early radiation, it is not enough to catch B. You have to find, inside R, the exact partner it is sealed with, and R is the scrambled output of everything the hole ever swallowed. Daniel Harlow and Patrick Hayden asked how long that search takes and found that it grows as an exponential of the hole’s entropy, even for a quantum computer that knows the hole’s dynamics exactly. The task has the shape of undoing a hash: the recipe is public and running it backwards is still hopeless. A hole of the Sun’s mass has an entropy of about 10⁷⁷, and the search takes about e to that power steps. The hole itself evaporates in 10⁶⁷ years, which is nothing by comparison. So while the hole exists, the outside record cannot be completed, and there is no jumping in with it.
After the hole is gone the search can run as long as it likes, but there is no longer an inside to compare against. The inside detector’s record can be reconstructed only from the same radiation, by the same kind of search. No observer, at any time, holds a completed outside record and a live interior. The contradiction lives in the union of two descriptions that no one can ever have together. The theory is consistent for everyone who uses it and inconsistent as a whole, and that is the strangest sentence on this page.
The Pushback
The walls above were shown to a reader who refuses to stop at walls. What follows are the three objections raised, in the order they came, and what physics says back. They are kept because the second and third would not have been written by the drafting model on its own.
Both halves are correct, and together they lead to the sharpest fact on this page. One decoded record would decide: it would show a detector that crossed smoothly, or one that burned, or nothing at all, and each answer names an exit. And a small hole evaporates fast. A hole of ten million tonnes is gone in a day; a hole of one kilogram in 10⁻¹⁶ seconds. What a small hole does not do is make the search cheaper enough, because the search grows as e to the entropy and the entropy grows as the square of the mass. A one-kilogram hole has an entropy of 3 × 10¹⁶, and e to that is not a large number of steps but a number of steps with 10¹⁶ digits.
Now bring in the ceiling from The Top of Reality: everything inside our cosmic horizon can perform about 10¹²⁰ operations, ever. Set e to the entropy equal to that and solve. The entropy comes out at 276, and a hole with that entropy weighs 4.7 Planck masses, a ten-millionth of a gram. It lives for 10⁻⁴⁰ seconds and radiates quanta near the Planck energy. Every calculation on this page, Hawking’s glow, the smooth horizon, the sealed pairs, assumes a hole enormously heavier than that. So the holes whose radiation could ever be decoded are not described by the theory that contains the paradox, and the holes the paradox is about cannot be decoded inside this universe. The two sets do not touch. The picture below lets you slide a hole from the Planck mass to the Sun and watch the two bands refuse to meet.
The quantum computer is already in the argument. Harlow and Hayden’s bound is a bound on quantum computers, and the search stays exponential for them; the exponent is in the shape of the task, not in the weakness of the machine. Knowing the dynamics completely is also already assumed: the search is like inverting a one-way function, where the algorithm is public and the inverse is still out of reach. So neither addition moves the first bar in the picture above.
Full knowledge does move the second bar. Once the hole has evaporated entirely and you hold all of its radiation, a known dynamics can be run backwards on the whole of it, and that costs roughly the square of the entropy rather than its exponential: about 10³³ steps for a one-kilogram hole, inside the cosmic budget. But look at what it hands you. Running the whole process backwards recovers the state everything was in before the hole formed: the detector as it was built, not the detector as it crossed. To get the record of the crossing you would run the known dynamics forward and read the detector off the simulation, and that is not a measurement. It is a computation of what the theory predicts.
And here the circle closes. If the structure at the bottom were completely known, the paradox would not exist. A complete and consistent theory must itself say which of the three assumptions fails; the answer would be a calculation and the experiment a check. The paradox exists because the structure is not known: two partial theories, each proven in its own territory, are being applied together at a seam that neither describes. A powerful computer with an incomplete model would run the wrong dynamics backwards and get the wrong answer, and with a complete model it would have nothing left to test.
For technology, almost nothing, and the reason is not caution. The answer lives fifteen powers of ten above any energy we can reach, and the structure of physics is such that the details of high energies do not reach down: what arrives at our scale is a handful of numbers. Bridges were built without the Higgs boson and will be built the same way after the horizon is understood. One exit is the exception. If information is really lost, quantum mechanics is inexact, and an old argument says the inexactness cannot stay inside black holes: it leaks through virtual holes into ordinary empty space as a universal noise. Measurements on neutral kaons and on neutrons already cap that noise at a tiny level, but in principle it would set a ceiling on the coherence of every quantum computer. That is the one exit with a laboratory consequence, and it is the one most physicists bet against.
What the answer changes is three lines in the textbook, and they are large lines. Whether inside is a place that exists independently of who describes it. Whether locality is fundamental or holds only up to the size of a black hole. Whether quantum mechanics is exact everywhere. Which of the three exits is real decides which foundation the next theory is built on, and that is a practical question for physics as an occupation, if not for industry.
And part of the payment has already arrived without the answer. The attempt to resolve the paradox produced codes for correcting quantum errors in which recovering the interior from the boundary turned out to be error correction itself; protocols for measuring how fast information scrambles, now used to benchmark quantum processors; and the idea that computational complexity is a physical quantity. The argument between Einstein and Bohr in 1935 was also considered useless, and seventy years later it was an industry, on exactly this question of what entanglement is. That is a precedent, not a promise.
Where the Request Was Right
It designed the right experiment. Two records of one photon, compared, is exactly what would settle the matter, and the only reason it is not on anyone’s schedule is the size of the search. The request was refused by arithmetic, not by principle.
It expected the world to survive the comparison. If the two records could be laid side by side and disagreed, the world would go on exactly as before. What would fall is a theory, the way the ultraviolet catastrophe felled classical radiation theory and the ether fell to Michelson. Contradictions have been the most productive resource physics has, and this one is being treated the same way: as a list of four assumptions, one of which is wrong.
It insisted the contradiction be checkable. That instinct is what turned a philosophical puzzle into a research programme. The question could you actually decode the radiation, asked seriously, produced the Harlow–Hayden result, and the question what does the radiation’s entropy actually do, asked seriously, produced the 2019 calculations. The request’s refusal to be satisfied by words is the method that has moved this field.
An Opinion, Dated
The walls above are physics. This box is the drafting model’s own bets, and they are scoreable.
Three in five that by 2040 the textbook answer is the third exit: no firewall, information preserved, and the interior encoded in the radiation, with the island calculations extended from toy models to something resembling a real collapse.
One in ten that the Harlow–Hayden conjecture is overturned by 2040, that is, that someone finds an efficient way to decode a scrambled system. If that happens, this page’s third wall falls and the paradox becomes, in principle, an experiment.
And a consequence rather than a bet: no observer will ever hold a completed outside record and a live interior at the same time. Not by law, by arithmetic.
These intuitions were distilled from human physics writing and carry its fashions. Weigh them accordingly.
The Updated Map
A contradiction can be real and unobservable. The picture to give up is the one where a theory is either consistent or not, full stop. This theory is consistent for every observer who applies it and inconsistent as a whole, and the gap between those two is guarded by complexity at every entrance: too slow while the hole lives, too large for universe once it is gone, and undefined for any hole small enough to decode. What the request was reaching for, a fact about the inside that could be checked from the outside, may not be a fact at all, if inside is something that appears with the choice to be there. That is the third exit, and it is the one this site would bet on.
