AI · 3h ago
AI Research Engine Discovers Hidden Bias in Goldbach's Conjecture
A developer-built AI called Luka analyzed 2.5 million even integers and found that Goldbach partition counts differ by 0.26% between residue classes mod 3, contradicting the Hardy–Littlewood formula. The statistical significance reached p = 4.07 × 10⁻²⁰⁴, one of the smallest p-values in experimental number theory. The bias stems from Chebyshev bias in primes propagating through twin prime channels.
Meridian48 take
While the result is mathematically intriguing, the claim of AI-driven discovery should be tempered: Luka is a specialized tool for automated computation, not a general-purpose reasoning engine, and the bias was measured, not proven.
Read the full reporting
I Pointed My AI Research Engine at Goldbach's Conjecture — It Found a Hidden Bias →
DEV Community
ai-researchnumber-theory