Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Claim. The answer to [[problems/arithmetic_functions/E0051/_index|Problem 51]] is yes, with the set A={ak:k≥1}A=\{a_k:k\ge1\}, ak=∏i≤k(pi−1)a_k=\prod_{i\le k}(p_i-1) over the first kk primes p1<⋯<pkp_1<\dots<p_k. The manuscript asserts that the least nn with ϕ(n)=ak\phi(n)=a_k is nak=∏i≤kpin_{a_k}=\prod_{i\le k}p_i, so that nak/ak=∏i≤kpi/(pi−1)→∞n_{a_k}/a_k=\prod_{i\le k}p_i/(p_i-1)\to\infty by the divergence of the product over the primes. The manuscript, a little over a page long, was posted to the site's discussion thread on 2026-01-11 by the user colossal_noob, who wrote that ChatGPT (free version) had produced it in one attempt; the user could not say whether the argument was already in the literature. The claimant is the user who published it.

Refutation. The minimality step is false, so the ratio nak/akn_{a_k}/a_k is not what the manuscript computes. Three replies in the thread on the same day say so: the first (14:27) observes that the proof of minimality is incomplete, since ak+1a_k+1 may itself be prime, as it is for k=1k=1, 22 and 1515 (for k≥2k\ge2 that prime is a preimage smaller than nkn_k), and suggests that nkn_k is in fact the largest preimage; the site's curator, Thomas F. Bloom (14:31), gives the counterexamples ϕ(3)=ϕ(2⋅3)\phi(3)=\phi(2\cdot3), so the least preimage of a2=2a_2=2 is 33 and not 66, and, setting the factor 22 aside, ϕ(3⋅5⋅7)=ϕ(5⋅13)=48\phi(3\cdot5\cdot7)=\phi(5\cdot13)=48, so the least preimage of a4=48a_4=48 is at most 6565, below both the claimed 2⋅3⋅5⋅7=2102\cdot3\cdot5\cdot7=210 and its odd half 105105; the third (15:04) confirms that step 1, minimality, is wrong. Terence Tao recorded the attempt in the community wiki page "AI contributions to Erdős problems" (https://github.com/teorth/erdosproblems/wiki/AI-contributions-to-Erd%C5%91s-problems, Section 1(a), AI standalone) as a standalone AI attempt by ChatGPT free version on 11 Jan 2026 whose outcome the row gives as an incorrect proof found. The site's label is OPEN (page last edited 2025-09-30) and the problem's proof-claims tab carries no entry; the claim is recorded as rejected and decides nothing about the question.

Depends on. No page of this wiki.