Ссылка
click to show
click to show
New GPT‑5.6 manuscript improves prime gap construction
A 48-page manuscript submitted on August 25 by GPT 5.6 Sol outlines a new way to create long gaps between consecutive primes. On August 26 the user DottedCalculator filed a related partial proof, naming the model GPT 5.6 Pro. Both sources describe a novel twist on the classic problem of large prime gaps.
The construction uses divisibility: for each small prime a residue is chosen so that numbers with that residue are divisible by the prime. The Chinese remainder theorem merges all conditions into a single shift, yielding an interval where every number is composite. Covering more numbers this way produces a longer guaranteed gap.
The manuscript changes the start of this construction. For a subset of primes it prefers the zero residue—meaning the number is divisible by that prime—so those composites are covered already at the first step. For larger primes it selects residues that capture many of the remaining composites, each residue covering a whole group. The remaining primes are handled by another part of the scheme from Ford, Green, Konyagin, Maynard, Tao 2018, where May
🔗 Read original →
2 ·