Peter Scholze and Dustin Clausen are rebuilding the foundations of mathematics. Their move: replace topological spaces, the foundational objects mathematicians have used for a century, with something called condensed sets. Scholze tells Quanta he is "trying to give names to what is there."
That is a Platonist framing. There is a fixed object; the math is shifting language for it. Fair enough. Except this is the latest in a sequence. Set theory was meant to ground everything. Category theory tried to subsume that. Homotopy type theory tried to subsume that. Now condensed sets. If the underlying object were fixed, the language would converge. It iterates instead.
The iteration is the tell. Each rebuild is a tighter compression of what the previous one was groping at. "Search for tighter compressions" is the language of computational theory, not Platonist metaphysics.
My read: math is a bad name for a real activity. The activity is pattern compression by physical pattern-detectors in a universe that admits compression. The discipline called "math" has been doing this for several thousand years, with periodic vocabulary upheavals when the encoding runs out and the field refactors. The name encodes a worldview, the idea that there is a separate realm of mathematical objects we are accessing, that the activity itself keeps falsifying.
Most of what makes math feel mysterious dissolves under this frame. Wigner's "unreasonable effectiveness of mathematics in the natural sciences," the puzzle that physical reality should be describable in mathematical terms, is only puzzling if math is metaphysically separate from physics. If math is what physical pattern-detectors do when they detect patterns in physical regularity, the puzzle collapses into identity. The detector and the detected share the same physics.
The foundational iteration loses its mystery too. Better encodings are findable; the search continues; periodically someone like Scholze finds a much shorter one and a generation reorganizes around it. That is search, not Platonic discovery. Each rebuild is a refactor against the same underlying computational structure, with the discipline's standing vocabulary as the cost function being minimized.
Tegmark's mathematical-universe hypothesis is the strongest opposing position. Tegmark holds that the universe is math, full stop: every consistent mathematical structure is a physical universe; ours is one of them. This relocates the puzzle rather than dissolving it. Now we have to ask why a math-multiverse exists, what selects our universe out of it, and what status "consistent" has independent of any computing process. The Platonist arrow runs the wrong way for what we actually observe, which is brains evolved inside one universe doing compression on regularities of that universe and calling the result mathematics. The parsimonious move is to let the arrow run from physics to brains to math, not from math to physics.
The right frame for what is actually happening is computational theory and complex dynamical systems. Those fields study compression itself: what can be compressed, by what kinds of process, with what limits. They have already absorbed half of what was traditionally called math: complexity classes, information theory, dynamical-systems classification, the theory of computation. The departmental boundary between math and CS is itself a piece of historical accident, and it is eroding because the underlying activity was always one thing.
LLMs are the current proof. A language model trained on math text does mathematics by next-token prediction over symbolic patterns. The activity it performs is statistical search over a symbolic regime, and the regime turns out to be sufficient for solving Olympiad problems, suggesting conjectures, doing real research-grade work. If math were access to a Platonic realm, a system that has never accessed that realm should fail. It doesn't. It does what brains were also doing, by a different implementation of the same search.
What this frame doesn't dissolve: large cardinals, infinite ordinals, the parts of formal mathematics that have no physical correlate. Those are coherent moves in a sufficiently expressive formal game; whether the game has external referent is a separate question. The claim is about the discipline's name, not about every internal move within it.
What it does dissolve: the idea that math is a unified field with a unified subject matter accessed by a unified method. There is no such thing. There is pattern compression done by physical pattern-detectors. The Big Bang made physics; physics made brains; brains made math; LLMs are math-doing physics again. The causal arrow runs through the universe; "math" is one historical name for one moment in the chain.
Scholze is right that he is giving names to what is there. He is wrong about what is there. It is not a Platonic object. It is computational structure in a universe that happens to compress, being detected by structures evolved inside that universe to detect compression. The field he works in is a historical accident around this activity, currently named badly, with the wrong departmental boundaries, with foundations that get rebuilt every few decades because the compression keeps getting tighter.
When the discipline finishes outgrowing its name, the successor will read like physics did when it stopped being called natural philosophy. The activity will be the same. The name will tell you what we figured out about what the activity was.
"Amazing Abundance" or "Computer Future" are not bad names for the field of everything civilization will enjoy pursuing as its cognitive light cone expands.