Erdős Problem #357
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: open (not automatically the live result).
Subject areas: number theory.
Read this status correctly
The status above is reproduced from GCL's immutable external-catalogue snapshot. It is not a new claim of current mathematical resolution, a source-semantic review, an independent proof check, or GCL certification. A published proof and a GCL replay are separate events.
The question in plain language
Choose a strictly increasing list of positive integers
Look at every nonempty consecutive block, such as \(a_2+a_3\) or \(a_1+a_2+a_3\). The requirement is that no two different blocks have the same sum. Let \(f(n)\) be the largest possible length \(k\). Erdős and Harzheim asked how quickly \(f(n)\) can grow, especially whether
In plain terms, the question asks whether the proportion of the interval \([1,n]\) used by such a collision-free increasing list must tend to zero. It is important that the terms are strictly increasing and that the sums are over contiguous blocks. Dropping monotonicity produces a different extremal function, and merely asking for many distinct sums is weaker than asking for every block sum to be distinct.
The original GCL assignment is issue #1627. Its source packet pins the catalogue snapshot and the formal declaration, but those pins are evidence for this audit, not a mathematical acceptance decision.
A small example
The list \((1,2,4,8)\) lies in \([1,8]\). Enumerating its ten nonempty consecutive blocks gives
These are pairwise distinct, so this exact calculation proves \(f(8)\geq4\). It is easy to reproduce with integer addition: start a running total at each position, extend the endpoint one term at a time, and insert each total into a set; the set has size \(10\), equal to the number \(4\cdot5/2\) of blocks. The same exhaustive checker gives, for \(n=1,\ldots,10\), the lower-and-upper maxima \(1,2,2,3,4,4,5,5,5,6\).
Neither calculation proves the asymptotic question. A finite list only supplies a witness, and exhaustive search only settles the tested values of \(n\). It cannot show that a construction exists for all large \(n\), or that every longer construction eventually fails.
What is known
The maintained Erdős Problems entry for #357 records the main question as open. It attributes to Erdős's 1977 survey the observation that an averaging argument forces \(a_k\gg k\log k\) for infinitely many \(k\) in an infinite analogue, hence zero lower density, and records the question about convergence of \(\sum 1/a_k\). The page also reports Desmond Weisenberg's lower bound \(f(n)\geq(2+o(1))\sqrt n\) via a related problem. Those are meaningful constraints, but neither answers the order of growth.
The page separately defines \(g(n)\) by allowing an arbitrary sequence in \([1,n]\), without strict monotonicity, and reports Hegyvári's 1986 bounds between \((1/3+o(1))n\) and \((2/3+o(1))n\). Hegyvári's primary paper is On consecutive sums in sequences, Acta Mathematica Hungarica 48 (1986), 193–200. The journal record was available, but the complete article was not available in this bounded source pass, so the exact theorem has been retained as a reported auxiliary result rather than independently certified here.
A useful adjacent primary source is Beker's 2023 paper on consecutive sums in strictly increasing sequences. Its Theorem 1.2 proves the existence of a strictly increasing sequence with at least \(c n^2\) distinct consecutive sums, for some constant \(c>0\). That does not solve #357: it permits repeated sums and therefore does not imply that all consecutive sums are distinct. The paper itself distinguishes this “many values” problem from the all-distinct extremal question.
The live page reports a further upper-bound attribution to Coppersmith and Phillips for an auxiliary formulation, while its discussion records a scope objection about whether that bound applies to \(f\) or \(g\). This audit does not use that disputed attribution as a theorem about the main problem.
GCL's current state
The frozen catalogue source is teorth/erdosproblems at commit 536408716724a2402bdfa7c1d764bf83d64a8c89, entry number: "357", blob 4e60263a80af6207aca494d2e1388e44fa372889. It records informal status open, with last update 2025-08-31, formal status unformalized, and statement-formalized metadata yes. The derived metadata also says formalized: yes with a later date; that internal date inconsistency is reported as metadata drift, not as evidence of a proof.
The pinned formal source is FormalConjectures/ErdosProblems/357.lean at commit 85f863718beeec7b58a3a1926ee92e3472bc2020, blob dd195174158df474e19c7ab3378c0073ce137e7f. Its definition uses strictly increasing maps Fin k → ℤ whose ranges lie in \([1,n]\), and its HasDistinctSums predicate is injectivity of sums over order-connected finite index sets. The principal theorem asks for \(f(n)=o(n)\), but its body is by sorry. The file also contains open and solved-labelled variants; those labels do not prove the main theorem or certify semantic equivalence with every website remark.
The original issue has zero comments at this audit, and no existing issue RESULT/1 was found. No GCL mathematical certification, accepted coverage decision, or protected research result is supplied by this article. V2 ledger state alone decides accepted coverage.
What remains
The smallest useful next step is a source-and-semantics reconciliation. First, obtain and read Hegyvári's full paper, recording the exact hypotheses and theorem location for the non-monotone \(g(n)\) result. Second, compare the website's “all sums” wording with the formal predicate, including whether the empty order-connected set is present and why positivity makes its zero sum harmless. Third, generate a compact exact certificate for \(f(n)\) on a fixed finite range, with a witness at the claimed maximum and a collision certificate for every longer candidate.
That work can clarify interfaces and expose transcription errors, but it will not settle \(f(n)=o(n)\). Any claimed asymptotic proof, counterexample, or Lean theorem replacement requires separate mathematical review and replay under the relevant authority.
Sources and claim boundaries
The original statement and historical source are Erdős and Graham's Old and new problems and results in combinatorial number theory, p. 58, together with Erdős's 1977 source cited there, p. 70. The maintained status and remarks are erdosproblems.com/357, whose live page was read through its indexed record and reports last edit 12 January 2026. The pinned catalogue and formal hashes are given above. Beker and Hegyvári are primary research references, but only Beker's accessible text was read in detail here.
The finite sums above are an independently run exact-arithmetic check, not an external proof. Published results, live-page reports, formal declarations containing sorry, unverified discussion comments, and GCL certification are separate evidence classes. The original SOURCE RESULT/1 citation is authenticated original RESULT/1. No future link is fabricated, and no status is promoted beyond the cited sources.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1627 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
- GCL triage: PROTECTED_READY_BACKLOG.
- Formal-statement coverage: A provider statement was included in the pinned snapshot. Presence does not mean the statement or proof has been checked.
- External mathematical status: consult current problem page and its dated bibliography.