Deterministic photon exchange prepared from the past

Two-ended control of photons — shaped emission and a time-reversed, matched receiver, both prepared in advance — is a mature technology: 94–99.4% coupling in single elements, 98.1% absorption of an arriving photon between remote nodes, links up to 30 m, and coherent memories that receive even unknown quantum states. Like the in and out states of an S-matrix element, both ends of the apparatus are prepared; this paper collects these realizations into one efficiency budget with exact shape and timing tolerances. It then asks what two-way quantum computing (2WQC) would need for the end of a computation to act as a condition, as preparation acts at its beginning. Everything we prepare — source, processor and a universal, problem-independent ready receiver to which a verifier routes a flag photon — fixes only the initial boundary of the Feynman ensemble. Standard quantum mechanics sums over the far end (an indifferent final boundary, giving the Born rule), so an acceptance condition can only be tested, with its ordinary probability. The missing ingredient is an enforced far end: a final boundary supplied by nature that weights ready endings. If weak but renewable, such a weight would compound exponentially over linearly many endings (about n·ln2/Δw), under conditions the paper states. We have not identified a published all-attempt bound for this weight. The paper specifies a direct test on existing hardware: two balanced endings, one active and one passive, selected after an earlier record is digitized and sealed. Standard quantum mechanics predicts exactly zero contrast, including every ordinary feedback path; 1e8 eligible trials per setting give a counting sensitivity Δw ≈ 3e-4 (planning figures: about an hour of link operation; a projected 1e-4 per delay on a quantum processor, with an analysis kit in the companion range-control paper). The test needs no absolute calibration: setting-independent preparation and readout errors cancel in the randomized difference. Either outcome constrains the declared ending law; a replicated signal would motivate tests of renewal, coherence and witness preservation. A one-page proposal for experimental groups is included.