Erdős Problem #289
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: open (not automatically the live result).
Subject areas: number theory, unit fractions.
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.
Editorial authority: The protected V2 admissions ledger alone determines publication status. This explanation is not independent mathematical certification.
The question in plain language
An interval of integers \([a,b]\) means the consecutive set
where \(a\) and \(b\) are positive integers with \(a\le b\). Its reciprocal weight is
and its length is \(b-a+1\). Erdős Problem #289 asks whether, once \(k\) is large enough, one can choose exactly \(k\) such blocks so that every block has length at least two, different blocks do not overlap, at least one integer is left unused between consecutive blocks, and the total reciprocal weight is exactly \(1\).
In symbols, the maintained modern formulation asks whether there is a threshold \(K\) such that for every integer \(k\ge K\) there are intervals \(I_1,\ldots,I_k\subset\mathbb N\) satisfying
with the blocks pairwise non-overlapping and non-adjacent. “For every \(k\ge K\)” is an eventual statement: it is stronger than finding examples for infinitely many values of \(k\).
There is an important historical qualification. On printed p. 34 of Old and New Problems and Results in Combinatorial Number Theory, Erdős and Graham define the notation \(\sum_{a,b}\) for the reciprocal sum from \(a\) through \(b\), and say it is “probably true” that, for sufficiently large \(k\), \(1\) can be written as a sum of \(k\) such block-sums with \(b_i>a_i\). The printed sentence does not itself say that the blocks must be distinct, disjoint, or separated by an unused integer. The stronger wording is the maintained catalogue’s natural restriction, not something that can silently be attributed to every word of the 1980 page.
A small example
The source gives a useful finite example for \(2\), rather than for the conjectured target \(1\):
Each block has at least two integers, and there is a gap between every pair. Exact rational arithmetic gives
and
I independently checked these fractions with Python’s exact Fraction arithmetic, including the lengths \(6,2,2,2,2\) and the strict separation condition \(b_i+1<a_{i+1}\). This verifies one finite identity and the intended geometry of separated blocks. It does not represent \(1\), does not cover every large \(k\), and gives no asymptotic proof. The example is attributed by the source to the solution of Monthly problem E2689, proposed by Hahn and associated with Montgomery’s 1979 publication; that attribution is not evidence that #289 itself is solved.
What is known
The directly inspected primary source is P. Erdős and R. L. Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), printed p. 34. Its wording is conjectural and less restrictive than the maintained formulation. On the same page it points to Hahn 1978 in connection with the proposed representation and to Montgomery 1979 for a representation of \(2\).
The cited Montgomery item is P. L. Montgomery, “E2689,” The American Mathematical Monthly 86 (1979), no. 3, p. 224, DOI 10.2307/2321534. The DOI metadata confirms the author, volume, year, page, and DOI. The source audit did not obtain a direct page image, so the precise attribution of the displayed five-block identity is retained as source-reported rather than presented as a fresh page-level transcription.
The live Erdős Problems entry currently displays the strengthened interval formulation as open and lists three unreviewed proof claims. Those are useful leads, not independent verification. One associated finite verifier reportedly passes arithmetic checks, but finite checks do not establish the eventual assertion for every sufficiently large \(k\). A repository or author manuscript can be evidence to inspect; it is not, by itself, a theorem certified by GCL.
GCL's current state
The authenticated source-audit record is original issue #1559, read under GitHub identity fyremael. Its exact RESULT/1 comment has disposition STATUS_CONFLICT. It reports the printed-source wording, the strengthened live formulation, the finite \(2\) example, the catalogue and formalization snapshots, and the unresolved source and replay gaps. This status must remain conspicuous: “STATUS_CONFLICT” is not a solved label.
The pinned catalogue packet identifies the source revision teorth/erdosproblems@536408716724a2402bdfa7c1d764bf83d64a8c89, entry number: 289, with Git blob 4e60263a80af6207aca494d2e1388e44fa372889. Its recorded catalogue status is open, its formal-solution status is unformalized, and its statement-formalized metadata is yes; the entry was dated 2025-08-31. The Programme registry binds issue #1559 to issue-body SHA-256 dd2424eb5cee8803a1f3ccc23eebdb977030502431a485d7e69b51fb526ee678 and classifies the item PROTECTED_READY_BACKLOG. These hashes establish provenance, not truth of the conjecture.
The audit also records a pinned Formal Conjectures statement at google-deepmind/formal-conjectures@85f863718beeec7b58a3a1926ee92e3472bc2020, blob a94483047e9448e90244569909e9fd8ce4f3df1d. Its separation predicate prevents overlap but permits adjacency, and it does not make positive endpoints explicit. A later provider snapshot uses a stricter non-adjacency predicate but still needs a semantic check against the intended positive-integer statement. The external proof source named by the audit is at commit ff33e501bb78a90ae4703e22e052f8bb38e10146, blob 547472bd9fdcc63acf5637f9db8ff6d2f26ce781; exact Lean replay was not completed because the required Lean 4.33.1 environment was unavailable. An external claimed Lean proof, author manuscript, or successful CI run is therefore not GCL mathematical certification.
What remains
First, a source adjudication must decide how the printed p. 34 formulation relates to the strengthened modern one. In particular, it must not silently turn an omitted condition into an original quotation. Second, the formal proposition must be compared line by line with the intended quantifiers: exactly \(k\) blocks, eventuality in \(k\), length at least two, exact reciprocal sum \(1\), non-overlap, non-adjacency, and positive endpoints.
Third, the external Lean artifact needs a clean replay at its exact source commit, followed by a dependency and axiom report. A successful compilation would still require checking that the theorem being compiled is the intended theorem and that no hidden placeholder or weaker predicate supplies the result. The positive-endpoint issue is especially concrete: if endpoints range over \(\mathbb N\) with \(0\) allowed, then the reciprocal convention at \(0\) must not quietly change the mathematical problem.
The bounded stopping condition is therefore clear: preserve STATUS_CONFLICT until the primary-source comparison, exact replay, dependency review, and endpoint bridge are all documented. The finite identity for \(2\) should remain a checked example, not be promoted into evidence for a claim covering every sufficiently large \(k\).
Sources and claim boundaries
- Erdős and Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), scan, printed p. 34; directly inspected scan SHA-256
0cbf0c32f0ab1e1c71db5121a88bac905bf976c4a6ab6bb6d7d9cf9ddd184ed3. - Erdős Problems #289, live catalogue wording and reported status; not a substitute for the primary source.
- Montgomery, “E2689,” The American Mathematical Monthly 86 (1979), p. 224, DOI and page confirmed in Crossref metadata; direct page-level verification remains the audit’s stated gap.
- Original GCL issue #1559, exact issue body and source packet; issue-body SHA-256
dd2424eb5cee8803a1f3ccc23eebdb977030502431a485d7e69b51fb526ee678. - Authenticated RESULT/1, disposition STATUS_CONFLICT, assignment
ERDOS-CATALOGUE-0289-S01; source audit and bounded successor design only.
This explanation reports documentary evidence and preserves its limitations. It does not claim that #289 is proved, that the external Lean source has been replayed, that the Monthly page has been freshly verified, or that GCL has admitted or certified the article. Publication and editorial admission are separate from mathematical proof status.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1559 — 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.