---

Claim. The answer to Problem 316 is no. Sándor's paper On a problem of Erdős (J. Number Theory 63 (1997), no. 2, 203--210) exhibits, as the site reports it, the set $D=\{2,3,4,5,6,8,10,12,15,20,24,30,40,60\}$ of divisors of $120$ other than $1$ and $120$: its reciprocal sum is $239/120<2$, and no partition $D=D_1\sqcup D_2$ has both $\sum_{n\in D_1}1/n<1$ and $\sum_{n\in D_2}1/n<1$. The site also reports his general theorem that for every $n\ge2$ some finite set $A\subseteq\mathbb N\setminus\{1\}$ has reciprocal sum below $n$ and cannot be split into $n$ parts all of reciprocal sum below $1$. The paper is not held by the library, so both statements are recorded second-hand from the site; the counterexample itself is a finite check, and the problem page records the corpus's exhaustive recomputation of it with exact rational arithmetic (all $2^{14}$ subsets of $D$), which confirms it. The strict inequalities are essential: $D$ does split into two parts with sums at most $1$, since $\{2,3,6\}$ has reciprocal sum exactly $1$.

Acceptance. Refereed: the paper appeared in the Journal of Number Theory, volume 63, issue 2, pages 203--210, in April 1997 (DOI 10.1006/jnth.1997.2113; the Crossref record gives the month without a day, and this page's date takes the first of that month). Reviewed: the site's curator, T. F. Bloom, marks the problem disproved and credits Sándor's counterexample; Bloom is not an author of the paper. The corpus has not read the paper; the acceptance rests on the refereed publication, the curator's credit, and the recomputed finite check.

Other counterexamples. The site's commentary also records the eleven-element set $\{2,3,4,5,6,7,10,11,13,14,15\}$, found by Tom Stobart, with reciprocal sum $120047/60060<2$ and no admissible split, and calls it minimal; the problem page records the corpus's recomputation of it and a ten-element set with the same property. Stobart's set appears only in the commentary, not in a dated posting, so it has no claim page of its own.

Formalization. The formal-conjectures file ErdosProblems/316.lean at the pinned commit (17 September 2026) states the question and proves the negative answer in the file by a decide +kernel step over Stobart's eleven-element set. Its docstring attributes the disproof to Sándor's paper and the formalization to Mehta, so it is linked above as a formalization of this result, with the witness changed from Sándor's set to Stobart's. Sándor's general $n$-part theorem is left as a sorry there. The corpus has not built or audited it, so no formalized evidence is listed.