For a prime p, let P (m) denote the largest prime factor of m and set κ (p) = P (p+1) − P (p−1) and S (x) = Σ≤ₗ κ (p). The analogous sum over all integers telescopes and is trivial; restricting the index to the primes removes the telescoping and makes S a genuine signed walk, whose growth is the subject of this note. We show that the walk's second moment satisfies D (x) = Σ≤ₗ κ (p) ² ≍ x³/ (log x) ², with the upper bound accessible by standard sieve methods (Brun–Titchmarsh) and the matching lower bound conditional on a Hardy–Littlewood prime-pair estimate; numerically, D (x) / (x³/log²x) is flat at ≈ 0. 143 up to 2×10⁶. Granting a decorrelation estimate for the increments together with a routine maximal inequality, the walk obeys the excursion law |S (x) | = x^3/2+o (1) /log x. We identify this decorrelation estimate, in substance, with the Erdős–Pomerance conjecture that the largest prime factors of neighbouring integers are independent — known to follow from the Elliott–Halberstam conjecture for friable integers — and the sign behaviour of κ with the Erdős–Turán conjecture on P (n) versus P (n+1). A surrogate test to 2×10⁶ places the real walk as an unremarkable member of the decorrelated null, with no dominant increment. The note proves no hard theorem unconditionally; its purpose is to define the object precisely, prove what is accessible, isolate where the difficulty lives, and place it correctly among these classical conjectures.
Samir Hanna Safar (Sun,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: