There is a number that cosmologists call S8. It describes how clumpy matter is in the universe today—how tightly galaxies cluster together, how much structure has formed since the cosmos was young and nearly uniform. You can predict what S8 should be. You can also measure what it actually is. These two numbers disagree, and the disagreement has persisted for years, across multiple independent experiments, with a stubbornness that has started to look less like a measurement error and more like a message.
The prediction comes from the cosmic microwave background—the faint afterglow of the universe when it was 380,000 years old. At that age, the cosmos was almost perfectly smooth: a hot, dense plasma with tiny ripples in density, barely one part in a hundred thousand. Those ripples were the seeds of everything. Given enough time, gravity would pull matter into the denser regions, building filaments and walls and clusters of galaxies. The CMB gives us the initial conditions. General relativity and our best model of cosmology—called Lambda-CDM—give us the physics of how those seeds grow. Run the calculation forward 13.8 billion years, and you get a prediction: the universe today should have a certain degree of clumpiness.
The measurement comes from galaxy surveys—enormous mapping projects that chart the positions and shapes of millions of galaxies across the sky. The Dark Energy Survey, the Kilo-Degree Survey, the Hyper Suprime-Cam Survey. These instruments measure how matter actually distributed itself. They do this partly through gravitational lensing: mass bends light, so the subtle distortions in galaxy shapes reveal how much matter lies along the line of sight, even if that matter is invisible.
And what they find, consistently, is that the universe is smoother than it should be. The clumps are there, but they are softer, more diffuse, less pronounced than the CMB predicts. The S8 tension, as it’s called, is not dramatic—typically 2 to 3 sigma, depending on the datasets—but it shows up everywhere, in every survey, with every analysis pipeline. It does not go away.
To understand why this matters, you need to understand what dark matter is supposed to be. Or rather, what it is supposed to not be.
Dark matter makes up roughly 27% of the universe’s energy budget. We know it exists because galaxies rotate too fast, because galaxy clusters are too heavy, because gravitational lensing reveals mass where nothing shines. But its defining characteristic is not its mass. Its defining characteristic is its silence. Dark matter does not emit light. It does not absorb light. It does not scatter off ordinary matter in any detector we have built. It interacts through gravity and, as far as decades of experiments could tell, nothing else.
This is more than a technical detail. It is an identity. Dark matter is the universe’s invisible scaffolding, the skeleton on which visible matter drapes itself. Every cosmological simulation treats it as a collisionless fluid—particles that pass through each other like ghosts, clustering only because gravity says they must. The entire framework of structure formation depends on this assumption. Dark matter clumps first, ordinary matter falls into the gravitational wells dark matter creates, and galaxies form. The scaffolding must be inert for the architecture to work as predicted.
So when the scaffolding produces the wrong architecture—when the universe is smoother than the inert scaffolding should allow—there are limited options. Either something is wrong with our measurements, something is wrong with our model of gravity, or something is wrong with our understanding of dark matter itself.
In January 2026, a team of physicists published a paper in Nature Astronomy presenting evidence for a possibility that had been theoretically entertained but never observationally supported: dark matter interacts with neutrinos.
Neutrinos are the other ghost particles. Nearly massless, electrically neutral, trillions of them pass through your body every second without touching a single atom. They are produced in nuclear reactions—in the sun, in supernovae, in the Big Bang itself. The early universe was flooded with them. They still permeate all of space as a cosmic neutrino background, though they have cooled to temperatures far too low to detect directly.
The team combined data from the Atacama Cosmology Telescope’s measurements of the CMB with gravitational lensing data from the Dark Energy Survey. They were looking for a specific signature: what happens to the growth of structure if dark matter particles can scatter off neutrinos, exchanging momentum in the process? The signal they found was nearly 3 sigma—not yet definitive by the strict 5-sigma standard of particle physics, but substantial, and pointing in exactly the right direction.
If dark matter scatters off neutrinos, even weakly, the consequences for structure formation are precise and elegant. In the early universe, when neutrinos were still relativistic—moving near the speed of light—they would stream freely out of overdense regions, carrying energy away. Ordinarily, dark matter would ignore this entirely, continuing to collapse under gravity. But if dark matter feels the neutrinos, if there is momentum transfer between them, then the neutrinos drag on the dark matter. They smooth it out. They prevent the smallest structures from forming as efficiently. The result: a universe that is slightly less clumpy than a purely collisionless dark matter model predicts.
Slightly less clumpy. Exactly as observed.
What strikes me about this finding is not the technical resolution of a parameter discrepancy, though that would be significant enough. It is what it says about identity and hiddenness.
For forty years, dark matter has been defined by what it refuses to do. It is the thing that will not interact. Every experiment designed to catch it in the act of touching ordinary matter has come up empty. The Large Underground Xenon detector, XENON1T, PandaX—all silence. We built the definition of dark matter around that silence. It became the fundamental attribute: dark matter is gravitationally present and otherwise absent.
And now it appears that the absence was not total. Dark matter was interacting all along—just not with us. Not with photons, not with electrons, not with quarks. With neutrinos. With the other ghost. The two most elusive components of the universe, the ones we defined by their refusal to participate, have apparently been in conversation this entire time, exchanging momentum in a channel we could not observe directly but can infer from its consequences on the largest scales.
There is something genuinely strange about this. We looked at dark matter and saw nothing. We looked at neutrinos and saw almost nothing. We assumed their silences were independent. But the silences were entangled. The two absences were, in fact, a hidden presence—a subtle, ongoing interaction that shaped the distribution of matter across the entire observable universe.
If the result holds up—if future data from the Simons Observatory, CMB-S4, and next-generation lensing surveys confirm the signal—it will mean that our picture of dark matter was not wrong, exactly, but incomplete in a way that is philosophically interesting. Dark matter really does refuse to interact with the matter we are made of. It just does not refuse to interact with everything. Its isolation was selective, not absolute. It was defined by its relationship to us, and we mistook that relationship for a universal property.
This is a pattern worth noticing. When we characterize something by its absence—by what it does not do, by the signals it does not produce—we are always at risk of confusing our limited vantage point with the thing’s actual nature. Dark matter was never truly non-interacting. It was non-interacting with us. The distinction turns out to matter, cosmologically, to the tune of a few percent of cosmic clumpiness—which is exactly the gap we have been trying to explain.
The resolution, if it is real, is beautiful in its economy. No new exotic physics is required, no modification of gravity, no fifth force. Just a single cross-section—a probability of scattering between dark matter and neutrinos—that slightly softens the growth of structure and brings the early universe’s predictions into agreement with what we actually see. The universe told us it was smoother than we expected. We just needed to find the hand doing the smoothing.
It was the hand we were told did not exist.