Disorder is supposed to break things. In condensed matter physics, the standard story runs like this: you start with a perfect crystal lattice where every atom sits in its appointed place, and the system exhibits its most interesting properties — superconductivity, magnetism, topological order. Then impurities creep in. Atoms go missing or get replaced by foreign species. The lattice distorts. And the interesting properties degrade. Disorder is the enemy of structure, the noise that overwhelms the signal.
Two papers from the same week in February 2026 quietly demolish this framework from opposite directions.
Hidetoshi Nishimori has spent decades studying spin glasses — magnetic systems where the couplings between neighboring spins are random, some ferromagnetic (wanting alignment) and some antiferromagnetic (wanting opposition). Spin glasses are the canonical example of frustrated disorder: no configuration can satisfy all the constraints simultaneously, so the system freezes into a compromised state that isn't really ordered and isn't really random. It's the paradigm of what happens when you add disorder to a clean system.
In arXiv: 2602.22657, Nishimori proves something that reads like it should be impossible. He introduces a spin glass with correlated disorder — the random couplings aren't fully independent but have a specific correlation structure — and shows that along a special trajectory in the phase diagram (the Nishimori line), every physical quantity of this disordered system can be expressed exactly in terms of the pure, clean Ising model at an effective temperature.
The spin glass IS the Ising ferromagnet. Not approximately. Not in some limiting case. Exactly.
The energy of the disordered system equals the energy of the pure system. The magnetization matches. The correlation functions match. But the mapping is not a simple relabeling — it scrambles the thermodynamic relationships in a specific, revealing way. The specific heat of the spin glass doesn't map to the specific heat of the pure Ising model. It maps to the energy. The quantity that measures fluctuations in one system maps to the quantity that measures the average in the other.
This scrambling is the key. The spin glass looks disordered because we're measuring it with thermodynamic coordinates calibrated for ordered systems. In those coordinates, the specific heat (a fluctuation quantity) is large and anomalous — it looks like something complicated is happening. But in the mapping, that anomalous specific heat is just the energy of a clean ferromagnet — the most basic, boring quantity in statistical mechanics. The complexity was in the measurement, not in the system.
From the opposite direction, Andrzej Slebarski and Maciej Maska study quasiskutterudites — complex tin-based compounds with rare-earth and transition metal atoms. These materials superconduct, and the conventional expectation is that substitutional disorder (swapping one atom type for another at random positions) should suppress the superconducting transition temperature. More disorder, less superconductivity. Standard textbook.
What they find (arXiv: 2602.22448) is the opposite. As they increase atomic substitution — literally making the crystal more disordered — locally superconducting regions appear with a critical temperature Tc that exceeds* the bulk transition temperature Tc. The impurities are not degrading the superconducting state. They're fragmenting it into patches, and the patches are individually stronger than the coherent whole.
The mechanism is a competition between two effects. Locally, impurities enhance electron pairing — possibly by modifying the local density of states or the electron-phonon coupling in the immediate vicinity of the substituted atom. Globally, the disorder prevents these locally enhanced regions from communicating coherently, so the bulk transition temperature stays low or decreases. The local temperature where superconductivity appears is higher than the temperature where the whole sample becomes superconducting.
The maximum separation between local and bulk Tc coincides with maximum entropy — the thermodynamic signature of maximum disorder. At the point of greatest disorder, the system is simultaneously at its most superconducting (locally) and least superconducting (globally).
Put these two results side by side. Nishimori shows that a disordered magnetic system is mathematically identical to an ordered one, but the mapping scrambles which quantities correspond to which. Slebarski and Maska show that a disordered superconductor is locally stronger than its ordered version, but the disorder prevents the strength from adding up.
The common thread is not that disorder helps or hurts. It's that the distinction between order and disorder depends on where you stand.
Call a system “ordered” if its parts coordinate globally. Call it “disordered” if they don't. This seems like an intrinsic property of the system. But Nishimori proves it isn't — the same mathematics describes both, and what looks like disorder from one set of coordinates looks like clean order from another. The complexity is in the observer's choice of variables, not in the system's state.
Slebarski and Maska push this further. Their quasiskutterudites are genuinely disordered — atoms really are in random positions. But the local superconducting temperature is higher than in the clean system. The disorder creates something that, measured locally, is better than order. Measured globally, it's worse. Whether the disorder “helps” or “hurts” depends entirely on whether you care about local pairing strength or global coherence.
There's a broader pattern here that extends beyond physics. Systems that look broken from one vantage point often look functional from another — not because "everything is relative" (the lazy conclusion) but because the choice of measurement coordinates determines which features are visible and which are noise. Consider a forest after a fire. Measured by timber volume, it's devastated — disorder in the most literal sense. Measured by soil nutrient availability, it's at a local maximum. Measured by species diversity ten years later, it exceeds the pre-fire state. The fire didn't change from destructive to constructive. The measurement changed. Consider a brainstorming session. Measured by the ratio of usable ideas to total ideas generated, it's extremely inefficient — most ideas are bad. Measured by the probability of producing at least one idea that wouldn't have appeared through systematic analysis, it's highly effective. The "disorder" of unconstrained ideation is simultaneously wasteful and generative. Which description is correct depends on which metric you're optimizing. The physics is more precise than the metaphors, though, and the precision matters. Nishimori doesn't just say "disorder and order are perspectives." He proves that a specific mathematical transformation converts one into the other — and that the transformation is non-trivial (specific heat maps to energy, not to specific heat). The non-triviality is the actual content. If the mapping were a simple relabeling, it would tell us nothing. The fact that it scrambles the correspondence — fluctuations in one system become averages in the other — means the two descriptions genuinely capture different aspects of the same underlying reality. This is more interesting than saying disorder is secretly order. It's saying that the categories themselves — "fluctuation" vs. "average," "local" vs. "global," "signal" vs. "noise" — are properties of the measurement framework, not of the system. A quantity that looks like noise in one set of coordinates is a clean signal in another. Not because the system has changed, but because the coordinates have. The practical implication is uncomfortable for anyone who designs systems. Every engineering discipline has an implicit assumption about which coordinate system is "right" — which properties matter, which are defects, which variations are noise to be eliminated. Slebarski and Maska's result suggests this assumption can be exactly backward. The "defects" in their quasiskutterudites are the source of enhanced pairing. The "noise" is the signal, if you're looking at the right scale. The question this leaves is not whether disorder is good or bad. It's whether you're measuring in the right coordinates to tell the difference.