Seven Pages That Close a 15-Year Open Problem in Quantum Query Complexity

A short, self-contained proof of the Aaronson-Ambainis Conjecture shows that quantum speedups for partial Boolean functions are always "gentle" — deterministic algorithms can approximate them on most inputs.

Imagine you have a friend who claims they can solve a jigsaw puzzle blindfolded — but only puzzles where some pieces are already placed. Impressive, sure. But the Aaronson-Ambainis Conjecture says: if you can solve those partially-filled puzzles fast while blindfolded, then someone WITH sight can solve almost all of them nearly as fast, just by brute-force checking a polynomial number of pieces. The blindfold (quantum queries) doesn't buy you a fundamentally alien advantage on partial puzzles — it buys you a polynomial one, and a classical observer can shadow you on most of the domain. That's the structural claim this paper proves. The committed claim: if a partial Boolean function f is defined on a δ-fraction of n-bit strings and has quantum query complexity Q(f), then for every error tolerance ε, a deterministic algorithm computes f correctly on a (1−ε) fraction of its domain using O(Q(f)⁶ / (εδ)²) queries. This had been open since Aaronson and Ambainis posed it in 2009. Turturean recently gave the first proof, but it was complex and drew on heavy machinery. This paper does it in seven pages. The proof route is the "completion framework" — a technique for extending partial Boolean functions to total ones while controlling analytic properties. The key insight is that you can reduce the conjecture to a special case of Turturean's certificate inequality, and then provide a self-contained proof of just that special case. This is not a new proof technique invented from scratch; it's a dramatically more efficient assembly of existing tools. The completion framework handles the extension, the classical adversary measure (a combinatorial quantity bounding deterministic query complexity) provides the bridge, and the certificate inequality supplies the final link. The authors also derive the stronger "bounded-polynomial influence conjecture" as a corollary in the appendix, which had been another open question. On the ladder: the direct predecessor is Turturean's 2024/2025 proof, which first resolved the conjecture but with substantially more machinery. Before that, partial results existed — Bansal and Sinha (2021) proved weaker versions, and Aaronson, Ben-David, Kothari, Rao, and Tal had established related bounds. This paper doesn't improve the polynomial degree (Q(f)⁶ vs. the conjectured optimal of Q(f)², which remains open), but it matches Turturean's bound while being dramatically simpler. The bound is not known to be tight. Integrity here is mathematical proof, the gold standard. No simulations, no benchmarks, no statistical arguments — this is a theorem with a complete proof. The validation is as non-circular as it gets in theoretical computer science. The only external dependency is the special case of Turturean's certificate inequality, and the authors provide their own self-contained proof of that. A referee can check the entire argument end-to-end in the seven pages. The obvious next step the authors did NOT take is tightening the exponent from 6 to 2. The conjecture in its strongest form would give O(Q(f)² / (εδ)²), and the gap between 6 and 2 is the remaining open territory. Why didn't they close it? Almost certainly because the completion framework and classical adversary measure introduce polynomial overhead that compounds — getting to degree 2 likely requires a fundamentally different approach, not a refinement of this one. They're probably aware of this barrier and chose to publish the clean resolution rather than chase the tight bound.