Strategic Resilience in the Post-Prime Era

Global digital trust — banking settlement, certificate authorities, command-and-control channels — is anchored in the assumed hardness of two operations: factoring large composites and computing discrete logarithms. Above that operational layer sits a second, quieter structure: an entire floor of conditional theorems in analytic number theory whose conclusions are stated 'assuming RH'. Prime-counting error terms, effective bounds on least primes, and much of the fine-grained arithmetic used in primality and key-generation analysis live on that floor.

That conditional floor is real, and it is worth naming plainly because it is the honest version of the 'house of cards' image. Working cryptographers already behave as though RH is true when they reason about the distribution of primes. A formal proof would convert hundreds of conditional statements into unconditional ones overnight — a genuine structural event in mathematics.

But the strategic brief has to survive its own audit here, so the correction goes in the text rather than in a footnote. RSA's security does not rest on the Riemann Hypothesis. RSA rests on the hardness of factoring a specific semiprime. RH describes the statistical regularity of primes in aggregate; it says nothing about how to split one particular product of two 1024-bit primes. Proving RH tomorrow would break no keys the day after. The single point of failure is factoring hardness, not prime regularity — and those are different objects that happen to share a vocabulary.

Keeping that distinction is not pedantry; it is what makes the rest of the argument usable. The risk in this essay is real, and it becomes visible only once the false version is cleared away.

The credible threat has never been the theorem. It is the toolkit built during the attempt. An adversary does not need a peer-reviewed proof to gain advantage; they need one intermediate lemma with unexpected computational teeth. This is the same pattern the archive has tracked elsewhere: the by-product exceeds the target.

Voronin's Universality Theorem (1975) is the sharpest illustration available. It establishes that the Riemann zeta function, restricted to a strip, can approximate any non-vanishing analytic function on a compact disc to arbitrary precision — the function is, in a specific technical sense, a universal simulator of analytic behaviour. The linear independence of the primes inside the Euler product is what buys that freedom. Zeta is also hypertranscendental: it satisfies no algebraic differential equation.

The strategic inference is that a new analytic machine capable of localising prime structure with far greater precision — the kind of machine a serious assault on RH would require — is exactly the class of object from which a factoring shortcut could fall out. The Guth–Maynard programme's translation of a zero-density question into matrices and largest-eigenvalue problems shows the pattern in miniature: the reformulation, not the conclusion, is where the new power lives.

Two guardrails. First, universality is an existence statement, not a construction: Voronin's theorem does not hand anyone an efficient algorithm, and nobody has turned it into one in fifty years. Second, 'could yield a shortcut' is a risk posture, not a result. The correct label for section 02 is contingency planning under uncertainty, which is precisely what a strategic function is for.

In statecraft, a breakthrough of this class is not a milestone; it is an unpublished capability. The precedent is documented rather than speculative: differential cryptanalysis was known inside the intelligence community years before its open rediscovery, and dual-use standards work has been shaped in ways later shown to favour the standard-setter. Any actor holding a factoring shortcut has an obvious incentive structure — harden internally on non-prime primitives, publish nothing, and leave counterparties on legacy RSA.

The result is what the brief calls the silent technological disadvantage: an organisation that believes itself secure while operating as an open book. It is the worst state to occupy because it produces no alarms. There is no breach notification for a decryption advantage.

The same logic already runs without any mathematical breakthrough at all, which is why the posture is justified today. Harvest-now-decrypt-later is a live, funded practice: intercept and store ciphertext now, decrypt when the capability arrives. Under that model the relevant question is not whether an adversary can read your traffic this year, but whether your data still matters in the year they can.

One correction to the brief's framing, in the archive's usual register: 'surrender of sovereignty' is the right stakes but the wrong causality. Failure to migrate is a failure of maintenance schedule, not an ordained fate. It is fixable by procurement.

Resilience means removing the shared dependency. Lattice-based cryptography — the Learning With Errors family, standardised as ML-KEM and ML-DSA — grounds hardness in geometric problems: finding short vectors, or decoding a point hidden in noise, inside a lattice of several hundred dimensions. Neither problem is a statement about primes, and neither yields to Shor's period-finding structure, which is what dismantles factoring and discrete logs.

That is the whole strategic point. A geometric trapdoor is orthogonal to the number-theoretic one. An unravelling of prime structure — however total — leaves a short-vector problem exactly as hard as it was, because the two rest on unrelated mathematics. Migration converts a correlated single point of failure into two independent ones, which is the only durable form of security engineering.

The costs are real and should be stated with the benefits. Lattice schemes carry larger keys and ciphertexts, they have roughly a decade of concentrated public scrutiny rather than four, and the field has already lost candidates late: SIKE was broken in 2022 after reaching a standardisation round, and Rainbow before it. This is an argument for hybrid deployment — classical and post-quantum key exchange composed together, so a break in either one alone is survivable — not an argument for delay.

Hilbert's thousand-year question is the right measure of the problem's weight and the wrong measure of an organisation's timetable. A strategic function cannot wait for an answer that has not arrived in a hundred and sixty-six years. It has to act on the structure of the exposure rather than on the arrival of the proof.

So the conclusion holds, in its disciplined form: migrate because the dependency is correlated, because harvest-now-decrypt-later is already running, and because the machinery risk — not the theorem — is the part nobody can bound in advance. Not because collapse is destined. A claim that arrives wearing inevitability has already stopped being an argument.

The honest version of the closing line: geometry is not the only architecture that survives, and nobody should promise that it does. It is the architecture whose survival does not depend on the same question. In a domain built on unproven assumptions, that independence is the entire asset.