Imagine you're in a massive library where every book is locked behind a librarian who will only answer yes or no to the question "Is this book in the collection I care about?" You need to find the book that overlaps with the most collections. A classical reader can only ask one librarian at a time, sequentially narrowing down. A quantum reader can ask all librarians simultaneously through superposition, then interfere the answers to amplify the best overlaps. This paper proves that the quantum reader doesn't just find the perfect book faster — it finds a better-than-classical book even when settling for "good enough." The committed claim: Decoded Quantum Interferometry (DQI) provably achieves a better approximation ratio than any polynomial-time classical algorithm on a specific oracle optimization problem called folded Optimal Polynomial Intersection (folded OPI). At code rate 0.3, DQI scores ~0.85 expected approximation, a modified variant scores ~0.95, while no classical algorithm can exceed the 0.65 threshold without super-polynomial query complexity. This is the first oracle separation for approximate optimization — previous quantum separations (Yamakawa-Zhandry 2022) only handled exact search. The architecture exploits a deep duality between optimization and coding theory. DQI works by encoding the optimization landscape into a quantum error-correcting code structure, running quantum interference across codewords, then decoding the measurement outcome back into an approximate solution. The key structural insight: the acceptance sets in folded OPI are constructed from folded Reed-Solomon codes, and the quantum algorithm's advantage comes from its ability to query membership oracles in superposition — something no classical algorithm can simulate efficiently. The modified DQI variant leverages recent improvements by Sun-Wootters, Horinaga-Yamakawa, and Jo to widen the gap further. On the ladder, this extends Yamakawa and Zhandry's 2022 exact-search oracle separation to the harder setting of approximation. The classical lower bound is proven via a reduction: any classical algorithm beating 0.65 on folded OPI would violate known query complexity bounds. The quantum upper bound comes from the DQI framework's connection to list-decodable codes. The gap is real but lives entirely in an oracle model — no physical quantum computer is running this today. The closest prior art is Jordan et al.'s DQI framework, which established the optimization-coding duality but did not prove a classical lower bound for approximation. Integrity is strong for a theory paper. The result is a mathematical proof, not a simulation or experiment. The classical lower bound extends established techniques (recording queries, statistical arguments over random oracle instances) to the approximation setting. The construction is explicit: folded Reed-Solomon codes with specific parameters. The numbers 0.85, 0.95, and 0.65 are derived analytically, not from Monte Carlo sampling. The 59-page paper provides full proofs. The main caveat: this is an oracle separation, which means the advantage is proven relative to a black-box oracle — it does not directly imply advantage on any specific natural optimization problem without further work. The milestone question is where this gets honest. Oracle separations are necessary but not sufficient for real-world quantum advantage in optimization. The next concrete target: demonstrating that the folded OPI structure (or something isomorphic to it) maps onto a natural combinatorial optimization problem — even an artificial-but-plausible one like MAX-k-CSP variants — without the oracle abstraction. If that mapping can be established, this framework becomes a serious contender for the first unconditional quantum advantage in optimization. Without it, this remains a powerful theoretical signpost. The experiment the authors did not run: instantiating a concrete (non-oracle) optimization problem where the DQI approximation gap survives. The honest read is option (c) — they're saving it. The paper explicitly notes the oracle setting and frames extending to natural problems as the key open direction. The 59-page proof infrastructure is clearly built to serve as scaffolding for that next step.
