Imagine you're designing a camouflage pattern. Your enemy has a library of every simple stripe, check, and gradient pattern — all the 'affine' designs. Your job is to create a pattern that looks as little as possible like ANY of them, no matter how they shift, rotate, or recolor. For decades, the best explicit camouflage designers could do was limited by a hard floor: no matter how clever your pattern, some stripe in the enemy's library would always match at least √N of your N pixels. This paper smashes through that floor. The committed claim: for any γ > 0, there exist explicit functions F: F2^n → F2^m with m = Oγ(n) such that agreement with every affine map is at most (1+γ)^n — exponentially smaller than the 2^{n/2} Fourier barrier that has bounded all previous explicit constructions. This is not an incremental improvement on an existing bound; it is a qualitative leap past a barrier that stood since the early 1990s. The prior-art ladder is well-defined. The Fourier-analytic method, rooted in bent functions and character sums from Nyberg (1991, 1993) through Carlet–Ding, Liu–Mesnager–Chen, Nagy, and Biryukov–Turecek–Udovenko, achieves agreement Θ(2^{n-m} + 2^{n/2}), which cannot go below 2^{n/2} regardless of output dimension m. For higher-degree polynomial agreement, Ben-Sasson and Kopparty (2010) used Gowers-norm arguments to get O(2^{-n/2^{d+1}} · 2^n). This paper beats both: (1+γ)^n for affine maps (degree 1) and the same for arbitrary degree d, with the output dimension scaling linearly in n. The improvement is super-polynomial. Architecturally, the method departs entirely from the Fourier/Gowers-norm toolbox that dominated this area. Instead, the authors establish a new bridge to classical algebraic geometry — specifically, results on counting solutions to systems of polynomial equations over finite fields. This translates the nonlinearity question into a combinatorial question about hypergraphs that simultaneously have few edges and small independent sets. It's a genuinely different proof technology applied to a problem that had been stuck in harmonic analysis for three decades. All results generalize from F2 to arbitrary Fq. Integrity here is the strongest kind available in pure mathematics: the results are proved by theorem, not simulation. The proofs are explicit and constructive — the functions can actually be written down, not just shown to exist probabilistically. The 29-page paper length suggests full proofs rather than sketched arguments. The baselines are clearly named with exact asymptotic expressions, and the improvement is stated precisely. There is no cherry-picking concern in proof-based work; the claim either follows from the argument or it does not. The milestone question in explicit nonlinearity construction is whether these exponential improvements can be made computationally efficient enough to impact practical cryptography — specifically, whether the Oγ(n) output dimension hides constants that make the constructions usable for realistic block cipher S-box sizes (n = 8 to 128). The current result is existential-to-explicit in the theory sense; the next real milestone is whether m/n ratios competitive with practical S-box parameters (say m = n, or m = n/2 for n = 128) can be achieved with concrete constants. The obvious experiment not run: instantiating these constructions at practical cryptographic parameters and benchmarking their actual nonlinearity profiles against the NIST lightweight crypto suite's S-boxes. The honest read is (a) — this is a theory paper from a theory group (Kopparty is a leading algebraic complexity theorist), and the computational implementation and cryptographic engineering are genuinely different skill sets. The algebraic-geometry bridge is the contribution; turning it into deployed S-boxes is a separate research program.
