Imagine you have a washing machine that, no matter what clothes you throw in, always produces the same shade of gray. You might think adding dirt (noise) would ruin any regularity. This paper shows the opposite: the dirt itself dictates the exact shade, and it's the same shade regardless of how the machine spins. The committed claim: when local quantum errors are present in scrambling dynamics, the late-time projected ensemble of mixed states converges to a universal distribution that depends only on the normalized Choi state of the error channel — not on the specific unitary dynamics underneath. This extends the concept of deep thermalization, previously established only for pure states, into the realistic noisy regime. Deep thermalization is the idea that sufficiently chaotic quantum dynamics produce projected ensembles (what you see when you measure part of a system and look at the rest) that converge to specific distributions over quantum states. For pure-state dynamics, this was shown to approach higher-order state designs. The open question was: does this universality survive real-world noise? Yu and Kos answer yes, and more — the noise itself becomes the fingerprint. The architecture is analytical quantum information theory applied to circuit models. They prove the result exactly for three progressively harder settings: dual-unitary circuits (the exactly solvable workhorse of quantum chaos theory), global Haar-random unitaries (the gold standard for random quantum dynamics), and local Haar-random circuits at large local Hilbert space dimension. The key mathematical object is the Choi state of the error channel, which encodes everything about the noise in a single density matrix. The projected ensemble's moments factorize in a way that pins them entirely to this Choi state. For dual-unitary circuits — the solvable case — they go further and show exponentially fast convergence to the limiting ensemble. This is not just an asymptotic statement; it tells you the result kicks in after a depth proportional to system size. Numerical verification covers qubit systems (local dimension 2, the hardest case for large-dimension proofs) and kicked Ising chains, both confirming the analytical predictions. The integrity picture is strong for a theory paper. The analytical results cover three distinct dynamical settings, preventing accusations of cherry-picking one tractable model. The numerics on qubit circuits and kicked Ising dynamics serve as genuine stress tests — these are not the settings where the proofs are easiest. However, everything is same-team calculation; no independent group has verified these results, and there's no experimental realization yet. What makes this more than a technical exercise is the implication for quantum simulation and error characterization. If the projected ensemble is determined entirely by the error channel, you could in principle read off the noise from the projected ensemble — turning a nuisance into a diagnostic. The paper doesn't develop this application, but the mathematical correspondence is exact and invites it.
