Imagine you run a courier service through a foggy valley. You know exactly how many packages you lose in one direction, and you've figured out the maximum throughput if you can send scouts back the other way to coordinate. But nobody could prove the theoretical ceiling — until now. This paper proves that when the fog only loses packages (pure loss, no added noise), letting scouts run both directions doesn't help beyond what a single clever strategy already achieves. The committed claim: the two-way quantum capacity and the two-way secret-key capacity of the pure-loss bosonic channel, subject to an average input photon-number constraint N, are both exactly g(N) − g((1−η)N), where g is the bosonic thermal entropy function and η is the channel transmissivity. This is a closed-form, single-letter formula. No optimization, no regularization, no asymptotic bounds that might tighten later — a final answer. Why does this matter? The pure-loss channel is the workhorse model for optical fiber and free-space quantum communication. Its one-way capacities (both quantum and private) were settled by Pirandola, Laurenza, Ottaviani, and Banchi in 2017 using the PLOB bound, building on earlier work by Takeoka, Guha, and Wilde. The two-way assisted capacities — where sender and receiver can use unlimited classical or quantum back-communication — remained open. The surprise here is that two-way assistance buys you nothing extra over the reverse coherent information protocol (sending half of a two-mode squeezed vacuum), which was already known to be achievable. The converse was the hard part. The architecture is information-theoretic, not computational. The proof works by establishing a matching upper bound (converse) to the already-known achievable rate. This places the result in the tradition of entropy-based capacity proofs going back to Shannon and, in the quantum setting, to Holevo, Schumacher, and Westmoreland. The key technical tool is bounding the two-way assisted rates using entropic continuity and properties of the bosonic thermal state. No simulation, no numerics — this is pen-and-paper mathematics. Integrity here is about mathematical proof, not benchmarks. A 13-page paper with one figure that derives a closed-form result lives or dies on the correctness of its inequalities and the tightness of its bounds. The achievability side was already established in prior literature (the reverse coherent information of a TMSV state); the new contribution is the converse. The validation is internal mathematical consistency, which in information theory is the gold standard — but it means the community will need to check the proof line by line. The milestone this completes is worth stating plainly: every capacity of the pure-loss bosonic channel is now known in closed form. One-way classical, one-way quantum, one-way private, two-way quantum, two-way secret-key — all done. What remains in bosonic channel capacity theory is the thermal-noise channel (pure loss plus additive Gaussian noise), where the minimum output entropy conjecture and its relatives still resist proof. That is the next frontier, and it is substantially harder because the thermal channel breaks the symmetries that make pure loss tractable. The obvious experiment not run — or rather, the obvious theorem not proved — is extending this to the thermal-noise channel. The authors almost certainly know this is the next target. The honest read is (c): this is the natural follow-up paper, and resolving the thermal case likely requires fundamentally new techniques beyond what works here. The pure-loss result is clean precisely because loss without noise preserves certain entropic structures. Adding noise breaks that, and the field has been stuck on it for over a decade.
