The Gap-3 Phenomenon
What Happens When You Actually Compute 13 Billion Cases
Most mathematical conjectures live in the land of “probably true.” This one didn’t get to stay there.
Over the past year, I’ve been working on a coding-theory result for a class of algebraic structures called finite chain rings — specifically, Z/p²Z, the integers modulo p-squared. These aren’t the standard Galois fields you see in classical coding theory. They’re rings: they have zero-divisors, a two-tier weight structure, and behavior that breaks the intuitions you build from fields.
The central question: what’s the maximum minimum distance you can achieve for circulant systematic codes at a specific parameter point? Not asymptotically. Exactly.
The answer turned out to be a hard ceiling. For every odd prime p — 3, 5, 7, 11, and beyond, the maximum is exactly 4. We call it the gap-3 phenomenon because the theoretical upper bound is 7 and the actual ceiling is 4. That gap of 3 is locked in, uniform across all odd primes.
We didn’t just prove it analytically. We computed it.
At p=5: 244,140,625 seeds evaluated.
At p=7: 13,841,287,201 seeds — every single one — zero exceptions to the ceiling.
Then we asked: Does it hold for two-power rings? Z/4Z, Z/16Z?
No. Z/16Z produces 3,072 seeds, breaking the ceiling. The gap-3 result is a property of odd-prime rings specifically. The two-power boundary is sharp and real.
---
Why does this matter outside of mathematics?
Because this result is the foundation of an architecture, the distance metric at the center of this paper is the same metric that drives a governance kernel I’ve built and deployed — a system that uses algebraic distance to evaluate whether a proposed AI state transition is consistent with a certified operational baseline.
The math isn’t decoration. It’s the enforcement mechanism.
Four provisional patents have been filed. The paper was submitted to IEEE Transactions on Information Theory today. The preprint is public now.
If you’re building AI systems and want governance backed by mathematical proof rather than a policy document, this is the direction.
Preprint → https://zenodo.org/records/20427913