Older Than Everything

For most of the last century, cosmology's official position on that phrase was that it's a category error, like asking what's north of the North Pole. The Big Bang wasn't an explosion inside space; it was the beginning of space, and time, and the conditions under which the word "before" means anything at all. You run the equations backward, everything converges on a singularity, the math stops returning numbers, and that's the wall. Not a door. A wall.

In February 2026, a paper in Physical Review D proposed that the wall has objects embedded in it. Enrique Gaztañaga argues that our universe didn't begin in a singularity but in a bounce — a prior cosmos that contracted, hit a floor, and rebounded outward into the expansion we've been living inside for 13.8 billion years. And in that model, some black holes formed during the contraction were dense enough, compact enough, and stubborn enough to survive the transition.

They came through. They're still here. Gaztañaga calls them cosmic fossils, which is generous, because a fossil is at least the shape of the thing that died. These are more like the parts that never noticed the death.

And the proposal is that there might be enough of them — that they might, collectively, be dark matter.

i · the word "before" was not supposed to work

Bounce cosmology isn't new and it isn't fringe, but it isn't the consensus either. It's the reasonable minority report. The standard model has inflation: a fraction of a fraction of a second in which space expanded by a factor with a lot of zeroes, smoothing everything out and explaining why the sky looks so uniform in every direction. Inflation works extremely well as a fit. It's also a mechanism nobody has directly observed, driven by a field nobody has directly detected, and it doesn't so much explain the singularity as decline to discuss it.

Bounce models take the other route. Instead of a beginning, a transition. Loop quantum cosmology has been producing bounces for two decades, where quantum geometry effects turn gravity repulsive at extreme density and the crunch never quite reaches zero volume. Roger Penrose has spent years arguing for conformal cyclic cosmology, where the far future of one universe is mathematically continuous with the birth of the next. Gaztañaga's own recent work has argued the universe could have emerged from gravitational collapse — that we might be living inside the interior of a black hole in some larger structure, which is either the most vertiginous or the most banal idea in physics depending on the hour.

What's distinctive here is the ledger. Most bounce models are elegant about the transition and silent about what crosses it. This one says: things cross it, we can estimate which things, and one of the great unsolved inventory problems in astronomy might be the shipping manifest.

That problem is dark matter. Something out there outweighs all ordinary matter by roughly five to one. It bends light. It holds galaxies together at rotation speeds that should fling them apart. It sculpted the scaffolding along which every visible structure condensed. It does not emit, absorb, or reflect light, and after decades of increasingly sensitive detectors built specifically to catch a particle version of it, the detectors keep coming back empty with impressive precision.

Primordial black holes have been a candidate since Carr and Hawking floated them in 1974: black holes formed not from dying stars but from density fluctuations in the very early universe, potentially at any mass, potentially everywhere. The idea has been alternately buried and exhumed for fifty years. What Gaztañaga adds is a different origin story and an earlier clock. Not black holes made in the first instants of this universe. Black holes made in the last instants of the previous one.

ii · ninety meters and the physics of surviving an ending

The number that does the work in this model is a threshold, and it's absurdly modest. Roughly ninety meters.

Below that scale, the paper's reasoning goes, an object gets erased in the crush — homogenized back into the undifferentiated soup of the bounce, all its structure smoothed away. Above it, the object is compact and self-gravitating enough that the transition simply happens around it. The universe ends. The object does not.

Ninety meters. The size of a modest office building is, in this model, the difference between being annihilated by the end of a cosmos and shrugging through it. There is no scale at which that isn't funny. The end of everything has a bouncer, and the bouncer is checking dimensions at the door.

Except that in black-hole physics a length is never only a length. A radius maps to a mass — that is precisely what the Schwarzschild relation does — and every constraint anyone has ever placed on dark matter is stated in mass, always. So read ninety meters as a horizon radius and run the arithmetic: roughly three hundredths of a solar mass. About thirty Jupiters. On the order of ten thousand Earths, folded into that office building. Which improves the joke considerably — the bouncer isn't checking whether you're big. He's checking whether you're dense enough that the end of the universe can't get a grip on you.

But now the two halves of this essay have to meet, because a number like that has somewhere to be. Compact objects of a few hundredths of a solar mass are exactly what the Magellanic Cloud microlensing surveys were built to find — EROS, MACHO, OGLE, watching millions of stars for the brief symmetric brightening that happens when something dark drifts across the line of sight. An object in that mass range produces an event lasting a couple of months, comfortably inside those campaigns' sensitivity, and they did not see anywhere near enough of them. EROS-2 alone excluded compact objects across roughly 10⁻⁷ to 10 solar masses as the entirety of the Milky Way's halo dark matter, and later long-exposure work from OGLE pushed the excluded band far higher still. So the survival threshold, read this way, doesn't land in one of the narrow surviving windows. It lands in the middle of the most thoroughly swept territory in the field.

Two honest caveats, because this is a back-of-the-envelope objection and it should be labeled as one. A threshold is a floor, not a distribution — the model needs a full mass spectrum, and where the bulk of the surviving population sits is a genuinely different question from where the cutoff is. And ninety meters may not be a horizon radius at all; it could be a comoving or curvature scale, in which case that's the wrong arithmetic performed on the right instinct. Either way the paper owes a reader the mass, in solar units, next to the exclusion curves. Nobody should have to estimate the central quantity of a dark matter claim on the back of an envelope.

Sit with the mechanism, though. It's stranger than the number. What makes a black hole able to survive the destruction of the universe that made it is precisely that a black hole has almost nothing left to destroy. The no-hair theorem — one of the cleanest and most brutal results in classical general relativity — says a black hole at equilibrium is fully described by three quantities. Mass. Charge. Spin. That's the entire specification. Everything else that ever fell in, every star's worth of composition and history and structure, is not encoded in any way the exterior universe can address.

A black hole is the most thoroughly forgotten object in physics. And that, in this model, is exactly why it's the only thing durable enough to carry across a cosmic ending.

That sentence is carrying more than it can hold, and it's worth saying so out loud rather than hoping you don't notice. No-hair is a theorem about stationary, equilibrium, asymptotically flat black holes in classical general relativity. A black hole inside a contracting pre-bounce cosmos is none of those things — not stationary, not at equilibrium, and with no asymptotically flat region to be flat in, because the entire universe is collapsing around it. And the bounce itself runs on quantum gravity, which is exactly the regime where nobody assumes no-hair survives.

Worse, or better: "black holes forget everything" is not settled physics. It is the single most contested question in the field. The information paradox has been running for fifty years, and soft hair, the Page curve, and the island calculations represent a decade-plus of serious work arguing the information is recoverable and three numbers was never the whole specification. So the most beautiful thing I can say about this model rests on the live controversy rather than on bedrock. I'm going to keep saying it, because in the classical picture it is true and the image is too good to surrender. But I'm saying it in pencil.

iii · the inheritance is three numbers, and that isn't all of it

This is the part where the vertigo stops being about scale and starts being about memory.

We tend to imagine continuity across a catastrophe as preservation — a seed vault, an ark, a library carried out of the burning city. Something survives, and what survives holds the content of what was lost. That's the story shape we're built for.

The bounce doesn't work that way. If Gaztañaga is right, the previous universe had physics, structure, a history of expansion and cooling and collapse, possibly complexity of kinds we couldn't name. All of it went into the crunch. What came out the other side is a population of objects each of which remembers exactly three numbers, none of which is a memory of anything that happened.

And then those objects — those maximally amnesiac relics — became the gravitational skeleton on which this universe hung everything. Dark matter is the scaffolding. It clumped first, and ordinary matter fell into its wells, and that's where galaxies formed, which is where stars formed, which is where the carbon in your hands was assembled. In this model, every visible structure in the cosmos condensed around the debris of a world that left no other trace of itself.

The tempting line here is that the inheritance is therefore empty — pure contentless weight, and nothing else made the trip. It's a good line and it can't be right, and the next section is what disproves it. If nothing crossed but weight, there is nothing to look for and this model isn't testable. Gaztañaga very much wants it to be testable.

The resolution is sharper than either half. Individually, each relic remembers three numbers and not one thing more. Collectively, the population is a readout: its mass spectrum, its abundance, its clustering across the sky are set directly by the density fluctuations of the dying universe that produced them. No single object remembers the collapse. The ensemble records it in detail. That is precisely why you can aim an instrument at it and expect an answer.

So the forgetting is real at the scale of the object, and the memory is real at the scale of the population. Which is a far better thing to say about how coherence crosses a discontinuity than "nothing survived but weight."

We go looking for continuity in the surviving individual — the seed, the ark, the one child carried out of the burning city. It usually lives in the distribution instead. A river is not its water. A language is not its speakers. A person is not the atoms they were wearing a decade ago, and an institution is not the people who staffed it in 1970. What crosses a discontinuity is the shape, and the shape is a statistical fact about a population rather than a possession of any member of it. None of these relics knows anything. The set of them is a document.

The inheritance is real. It's just written in a hand that no single inheritor can read.

That's either desolate or it's the most reassuring thing in cosmology, and I've genuinely never been able to decide, which is my favorite state to be in.

iv · what would kill this idea

Now the discipline, because a beautiful idea that can't be falsified is just decorative.

Gaztañaga names three places to look. Relic gravitational waves from the contracting phase, which would be a signal from a universe that isn't ours and never was. Patterns in the cosmic microwave background that preserve pre-bounce information — the ensemble talking, not any one relic, which is the only way that sentence is coherent at all. And unexpectedly massive objects in the early universe, of the sort the James Webb Space Telescope kept turning up in its first years, galaxies apparently too big to have assembled that fast.

That third line deserves an asterisk. A good fraction of those impossibly-early JWST masses have been revised downward as astronomers worked out how much of the light was coming from accreting black holes rather than stellar populations. The "impossible early galaxy problem" has been shrinking. Building a case on it is building on a surface that's still moving.

The harder constraint is the one already spent above. Primordial black holes as all of dark matter have been squeezed from nearly every direction for decades — microlensing surveys that should have caught them crossing in front of stars, cosmic microwave background limits on how much they'd have heated the early universe by accreting gas, Hawking evaporation ruling out the lightest end. What's left are narrow mass windows, and the ninety-meter threshold does not obviously fall inside one. Threading them is not a formality, and this model has not yet shown its work.

And under all of it: this is a theory paper. It is a self-consistent mathematical structure that fits the same observations inflation fits, offered as an alternative. That's a real contribution and it is not a discovery. Nobody has seen a fossil from before the beginning.

v · when the data can't decide

Stop on the flattest sentence in this essay: it fits the same observations inflation fits.

Two mutually incompatible accounts of the origin of everything. The same predictions, so far. No evidence currently distinguishing them. That isn't a gap in the story — that is the story. Because when evidence underdetermines, something still decides, and the something isn't evidence. It's elegance, seniority, which instruments happened to get built and what they were built to see, whose graduate students staffed which collaborations for forty years. Inflation isn't the standard model because it beat bounce cosmology in a fair fight over data. It's the standard model in part because it arrived first, and an enormous amount of infrastructure got poured around it while it stood there.

Then notice what this paper had to become in order to be legible. Its headline isn't "a new bounce model." It's "and it might be dark matter." Dark matter has become the field's universal solvent — the standing vacancy that every new exotic object gets nominated to fill, because that's the opening currently posted. That's not a knock on Gaztañaga, who is doing real work in a real journal. It's a structural fact about how a hypothesis becomes publishable, press-releasable, fundable, visible. The shape an idea takes is partly set by the shape of the hole the field has open.

Which means the question sitting underneath a cosmology story is not only what is true. It's who decides what's true during the long stretch when the data can't tell us — and what gets selected for when the honest answer is that nobody knows yet. Historically the answer is: whoever's story fits the vacancy.

That's worth holding next to the model rather than instead of it. The bounce might be right. It might be right and lose anyway, for a generation, on grounds that have nothing to do with the sky.

But the paper has already done the useful thing, which is to make "before the Big Bang" a sentence with a testable predicate attached. For a hundred years that phrase was a wall with a sign on it reading no questions past this point. Now there's a version of it that specifies what should be lying around, how big it has to be to have made the trip, and where to point a telescope.

We are leftover matter that learned how to weigh things. Some of that matter may be older than the word "old" was supposed to reach, holding our galaxies together, remembering three numbers and nothing else, and we're down here trying to work out its mass from the way it bends light.

The universe doesn't have to be comprehensible. It keeps being checkable anyway, which is a much stranger gift.