Imagine you're playing Wordle, but instead of guessing one word, you have a list of candidates for each letter position, and you're trying to find a single word that hits as many lists as possible. Now replace 'word' with 'low-degree polynomial over a finite field' and 'letter positions' with 'evaluation points,' and you have the Optimal Polynomial Intersection (OPI) problem. This paper attacks OPI and its cryptographic twin — local leakage resilience (LLR) for Shamir secret sharing — and moves both needles in ways the field thought were stuck. The committed claim has two parts. First, Shamir secret sharing is one-bit leakage resilient for rate R = k/n ≥ 1/2 − δ for some constant δ > 0. This smashes through the so-called 'one-half barrier' that had stood as the frontier since Benhamouda, Degwekar, Ishai, and Rabin's 2018 work, and dramatically improves over the previous best of R ≥ 0.668 due to Kasser (2025). Second, they give a quantum algorithm for OPI that matches the information-theoretic optimum (the semicircle law of Jordan et al., 2025) up to ε, beating prior algorithmic AND existential results. The architecture is algebraic and combinatorial at its core, leveraging a recent structural bridge the same authors built (Sun & Wootters, 2026) connecting LLR and OPI. The quantum component is not variational or circuit-heavy — it exploits quantum query complexity to achieve a provable separation between quantum and classical hardness for OPI over large fields, adapting a framework from Yamakawa and Zhandry (2024). The classical side is pure math: Fourier analysis over finite fields, character sums, and entropy arguments. No GPUs, no training — this is pencil-and-paper theory with a quantum oracle result layered on top. The ladder here is clear. For LLR, the named baseline is Kasser 2025 (R ≥ 0.668); this paper pushes to R ≥ 0.5 − δ. For OPI algorithms, the baselines are Jo (2026) and Horinaga & Yamakawa (2026); this paper's quantum algorithm achieves the semicircle-law optimum, which those prior works fell short of both algorithmically and existentially. The quantum-vs-classical separation is unconditional relative to membership oracles, which is strong for a complexity-theoretic result. Integrity is high for a theory paper. The results are mathematical proofs, not simulations or experiments — the validation is internal consistency and logical correctness, not benchmarks. The 70-page length suggests the proofs are detailed rather than sketched. The comparison baselines are current (2024–2026 citations), not stale. There's no cherry-picking risk in the empirical sense because there are no empirical results — but the choice to state the LLR threshold as 1/2 − δ (for unspecified δ) rather than a concrete number is worth noting. The paper does not pin down how far below 1/2 the rate can go. The obvious next milestone is pinning down the exact constant δ — or, more ambitiously, pushing all the way to R > 0 (or some much smaller threshold). For OPI, the milestone is dequantizing the quantum algorithm or proving it cannot be dequantized. The quantum-classical separation is relative to an oracle; removing the oracle would be a major complexity-theory result. The experiment not run: explicit computation of δ. The paper proves existence of a constant δ > 0 but does not optimize it. Honest read: this is a structural limitation of the proof technique (likely character-sum bounds), not compute budget. Optimizing δ is probably the next paper — either by these authors or by someone who tightens the Fourier-analytic estimates.
