Imagine you're a mail carrier who has to deliver two sealed envelopes to two neighbors living in the same duplex — but each neighbor is also a nosy spy trying to read the other's mail, even if they already know the contents of their own envelope. The carrier's only tool is a single, shared hallway (the broadcast channel). The question: how much secret mail can you deliver to each neighbor simultaneously? This paper answers that question for quantum hallways. The committed claim: for broadcast channels with classical inputs and quantum outputs, the authors establish the first Marton-type achievable rate region for two independent confidential messages under the strongest meaningful secrecy criterion (conditional strong secrecy). This isn't a single-receiver wiretap channel — those have been studied extensively since Wyner (1975) and Devetak (2005). This is the strictly harder two-receiver, two-secret problem where the channel itself is quantum. The core machinery is a bipartite classical-quantum resolvability theorem. Standard resolvability says you can approximate a target output distribution by choosing codewords at random from a codebook. The twist here is that the encoder must select pairs of codewords — one for each receiver — using normalized likelihood weights, and this pair-selection creates statistical dependence that threatens to leak information. The authors show that even with this dependence, the output on each receiver's side can be made indistinguishable from the 'no-message' distribution, killing any eavesdropping advantage. The proof chains a change-of-distribution argument for reliability with the new resolvability result for secrecy, all within the same encoder construction. The ladder context matters. For a single confidential message over a quantum wiretap channel, Cai-Winter-Yeung (2004) and Devetak (2005) established the capacity. For classical broadcast channels with confidential messages, Liu-Maric-Spasojevic-Yates (2008) gave the Marton inner bound. This paper is the quantum generalization of the latter, and it recovers known results for deterministic and degraded cases as special instances. The multi-letter capacity characterization they derive is exact in principle but not single-letter (a common situation in network quantum information theory where computable formulas remain open). The most striking concrete result: for the coherent isometric extension of the classical Blackwell channel, the authors compute the full confidential classical capacity region AND the unassisted quantum capacity region — then show that certain confidential classical rate pairs lie outside the quantum capacity region. In plain terms: there are communication tasks where sending classical secrets beats sending quantum information, even on a quantum channel. The Platypus channel provides a second example of this separation, confirming it isn't an artifact of one special channel. Integrity is grounded in mathematical proof rather than simulation or experiment — appropriate for an information-theoretic result. The proofs are self-contained (change-of-distribution for reliability, bipartite resolvability for secrecy), and the explicit evaluations for the Blackwell channel provide checkable numerics. No pre-registration or code release applies here — this is theorem-proof territory, not computational science. The main vulnerability is that multi-letter characterizations resist single-letter simplification, leaving the full capacity region uncomputable in general. The successor question is transparent: the obvious next step is a computable single-letter capacity formula or a matching converse. The authors likely didn't push this because converses in quantum network information theory are notoriously hard — superadditivity of quantum information measures means the standard classical toolkit (strong converse, tensorization) breaks. They're probably not hiding a failed attempt; this is a known open wall in the field.
