The standard resolution of Maxwell’s demon goes like this: the demon measures a molecule’s velocity, sorts it into the fast or slow compartment, and appears to decrease entropy for free. But eventually the demon’s memory fills up and must be erased. Landauer’s principle says erasure costs at least kT ln 2 of energy per bit. This cost pays back the entropy debt, saving the second law. The villain is erasure.

Kastner (arXiv:2503.18186) argues the villain is measurement. The entropy cost attributed to erasure actually originates in the act of localizing the molecule’s state — the measurement itself carries the thermodynamic penalty, and it does so for a reason grounded in quantum mechanics rather than in information theory.

The argument traces back to Brillouin’s 1951 position, which Bennett’s 1982 paper claimed to supersede. Bennett argued that measurement could be performed reversibly — at zero entropy cost — pushing the entire thermodynamic burden onto erasure. Kastner contends that Bennett’s reversible measurement relies on unphysical classical idealizations. In any physical realization, localizing a particle to a specific observable value requires resolving it against quantum uncertainty, and this resolution is thermodynamically costly. The cost is not optional; it’s baked into the physics of acquiring information about a quantum system.

The through-claim is about where the debt is incurred. The final number — kT ln 2 per bit — is the same regardless of whether you attribute it to measurement or erasure. The thermodynamic accounting balances either way. What changes is the physical story: in Bennett’s version, the demon operates for free until cleanup time; in Kastner’s version, the demon pays at the door. The same charge, different billing cycle.

This matters because the causal story determines what you think the fundamental constraint is. If erasure is the cost, then information is free to acquire and expensive to discard — a universe where knowledge is cheap and forgetting is hard. If measurement is the cost, then information is expensive to acquire and the cost of discarding it is derivative — a universe where knowing is the hard part. The same thermodynamics, but two different accounts of what’s difficult about understanding the world.