Wiki
Wiki

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

Updated

Library

additive_bases/: Sources filed under Additive Bases and Sidon Sets, with cross-references to related work.

additive_combinatorics/: Sources filed under Sumsets and Arithmetic Progressions, with cross-references to related work.

analysis/: Sources filed under Analysis, with cross-references to related work.

arithmetic_functions/: Sources filed under Arithmetic Functions, with cross-references to related work.

covering_systems/: Sources filed under Covering Systems, with cross-references to related work.

diophantine_problems/: Sources filed under Diophantine Problems and Powers, with cross-references to related work.

discrepancy/: Sources filed under Discrepancy, with cross-references to related work.

discrete_geometry/: Sources filed under Discrete and Convex Geometry, with cross-references to related work.

distance_problems/: Sources filed under Distance Problems, with cross-references to related work.

divisors/: Sources filed under Divisors and Multiples, with cross-references to related work.

extremal_graph_theory/: Sources filed under Extremal and Structural Graph Theory, with cross-references to related work.

factorials_binomials/: Sources filed under Factorials and Binomial Coefficients, with cross-references to related work.

graph_coloring/: Sources filed under Graph Coloring, with cross-references to related work.

group_theory/: Sources filed under Group Theory, with cross-references to related work.

integer_sequences/: Sources filed under Sequences and Densities of Integers, with cross-references to related work.

irrationality/: Sources filed under Irrationality and Diophantine Approximation, with cross-references to related work.

number_theory/: Sources filed under Other Number Theory, with cross-references to related work.

polynomials/: Sources filed under Polynomials, with cross-references to related work.

primes/: Sources filed under Primes, with cross-references to related work.

ramsey_theory/: Sources filed under Ramsey Theory, with cross-references to related work.

set_systems/: Sources filed under Set Systems, Designs and Hypergraphs, with cross-references to related work.

set_theory/: Sources filed under Set Theory and Infinite Combinatorics, with cross-references to related work.

unit_fractions/: Sources filed under Unit Fractions, with cross-references to related work.


The literature, filed under the same subject categories as the problems. Each source has one primary home at <subject>/author_year_slug/, with cross-references from other relevant categories. Its folder holds an _index.md whose desc is the catalog entry and whose body is the digest — what the paper does, its method, and which problems it touches — and one page per extracted result, named by the paper's own label (theorem_1, lemma_2_3, corollary_4; conjecture_p30 for an unnumbered statement on page 30; a descriptive name such as main_theorem when the paper gives none), and the source's file when the library may hold it. A result page carries a title:, a one-sentence desc, the precise statement, a pointer to the proof or a sketch, the results it depends on, and a "Bears on" list linking the problem pages it concerns. Result pages, digests and Bears-on rows are written in the corpus's own words: the statement restated with its hypotheses intact, the proof pointed to or sketched here, a quotation kept short, attributed and used only where the exact wording matters. The source's text lives only in the PDF and in a transcription beside it. The source is the PDF under the folder's name when one exists; a markdown transcription may sit beside it as author_year_slug.md. When no PDF exists — a web page, a forum answer — that markdown file is the source itself and records the URL. Where a PDF exists it is canonical: result pages cite it by page and theorem number, and when a transcription disagrees with it, the PDF wins. Held PDFs and files over 1 MB live in Git LFS, whose pointer records each file's size and SHA-256; a smaller held file is identified by its path. The card's provenance line names which edition the held file is (a publisher PDF, an author preprint or a scan, with its URL and retrieval date where known); when the held edition differs from the one a problem page cites, the card says so and result pages cite the held edition's pages. The repository holds a paper's file only under an open license; otherwise the card cites the source, names the version read and says that no file is held because no license on record permits redistribution. Each card records the terms it observed for every file it holds or read. The frontmatter key license carries one term when the folder holds one file and a mapping from file name to term, keys in byte order, when it holds several; a card holding no file may still carry the term it read. A term is an SPDX license identifier as the SPDX list spells it (CC-BY-4.0, Apache-2.0), a Creative Commons license whose version the source does not state, named without one (LicenseRef-CC-BY, LicenseRef-CC-BY-NC-ND), reserved (every right reserved: a copyright or usage notice, a publisher's or repository's non-exclusive distribution license, or arXiv's assumed license), or unstated (no terms found on the file or in its record). When sources disagree, an explicitly named open license on the rights holder's page for the held edition decides, and the card records the older notice beside it with the date the page was read; otherwise the term comes from the arXiv record for an arXiv file, then the notice the file prints, then the publisher's page, then a web source's terms. "Free to read" without a named license is not a license, and where nothing in that order yields a term (the work's page cannot be read and no other source was observed), the term is unstated, with the reason. The card's provenance text says where each term was read -- the notice the file prints, quoted as printed; the publisher's or repository's record, with its URL and the date it was read; or an archive's own license file -- infers nothing beyond it, and claims no redistribution right the record does not establish. erdos license-audit checks the key against the held files and the arXiv records (docs/tools.md "License audit contract").