Plectis

Current papers

Current papers

Rendered from paper/README.md in the public Lean repository. View source

Papers

Each of the eight covered Erdős problems has a short first-read paper and a longer complete reasoning record. Start with the short paper for the result, its argument and the question that remains. Use the longer record when you want to recover a detailed step, inspect a computation, or follow an approach that stopped. You do not need Lean or a coding agent to read either.

Using the degree-seven polynomial constructed by the erdosproblems.com contributor ani, Lean proves that every preconnected strict-lemniscate set containing two distinct roots has one-dimensional Hausdorff measure greater than two. This refutes the exact Formal Conjectures path-image-length statement; the separate total-variation bound is also checked. The other seven targets remain open. Independent human review of correspondence with the 1958 wording has not been recorded. Comparator checks only selected exact statements, axioms and kernel acceptance; it does not assess novelty or historical correspondence. The papers report partial results, failed or equivalent routes, finite evidence, and the exact obligations that survive. If the problem numbers or formalisation are unfamiliar, read a reader’s way in first. For the design of the tools and the research process, go to the project papers.

Problem papers

Problem Short paper Complete reasoning record
#68 Integer Linear Forms for a Factorial Reciprocal Series (source) Factorial Linear Forms and Denominators: Detailed Proofs and Rationality Criteria (source)
#243 Cubic-Rate Irrationality and Reciprocal-Tail Rigidity (source) Reciprocal-Tail Rigidity: Theorems, Proofs and Questions (source)
#249 Integral Relations among Totient Sections (source) The Binary Totient Series (source)
#251 Sparse Congruence-Preserving Perturbations of Dyadic Series (source) Prime-Gap Dyadic Series: Perturbations, Exact Criteria and Certificates (source)
#257 Weighted and Covered Supports for Mersenne Subseries (source) Reciprocal Mersenne Subseries (source)
#269 Distinct running least common multiples (source) Running least common multiples: distinct heights and repeated sums (source)
#1041 Paths in Polynomial Lemniscates: A Degree-Seven Counterexample and Radial Connections (source) Paths in Polynomial Lemniscates: A Degree-Seven Counterexample and Radial Connections (source)
#1049 Hankel Determinants of Geometric Moments and Rational Lambert Values (source) Geometric Moments and Rational Lambert Values: Proofs and Further Results (source)

The #243 short paper leads with irrationality under the cubic rate an2/an+1=1+3/n+o(n−3) for strictly increasing positive integer sequences. Its bounded-increment criterion for an eventual Sylvester tail is a second result; the unrestricted problem remains open. The #1049 short paper proves a rational-base sufficient region using Zudilin’s forms after cyclotomic cancellation and denominator accounting. Its region contains 31/4, which is outside the earlier Bundschuh–Väänänen sufficient region; 3/2 remains open.

Reading the eight together

One cross-problem paper develops the mathematics that arises from reading the programmes together. For c>0 and integers d≥1, it considers entire functions f(z)=∑n≥1enzn/n! with nonnegative integer en eventually bounded by nc and, for each fixed positive integer q, eventually divisible by q. The set of derivative-value vectors at 1 has Hausdorff dimension min(c,d) and contains an open set exactly when c>d. This result has an ordinary proof with the formal ingredients identified separately. The paper also contains the capacity and congruence criterion, Lambert-subsum results, exact computations and method limits.

Reading Eight Erdős Problems Together (source).

Follow the argument into its evidence

Read the assumptions of the result you want to use, then the proof. The longer record gives the surrounding working context, including routes that failed, finite experiments and open obligations. If you cannot reconstruct a step or find its source, that is useful feedback through Contributing.

Some results are checked in Lean; others are ordinary mathematical arguments or applications of cited external theorems. Each paper states the boundary for its own claims. Results and limits gives the result beside what remains open, and the source map connects the paper to supporting declarations. Lean files are proof authority only for the exact declarations they check; docs/claims.json records the selected public claim interfaces.

The full-text index provides generated Markdown versions for browsing. The PDFs and .tex files above are the authored manuscripts. The older joint #249/#257 paper is retained for provenance; it is not the entry point for either problem.

Project papers

Start with A Repository-Based System for Research and Publication (source). This is the main systems paper: it follows a result from its mathematical argument through formal support, written explanation, review and contribution.

Two earlier papers are retained as historical background. Their account is superseded by the main paper; their observations and cross-references belong to the revisions they describe.

Earlier paper Detail retained
From a Cold Clone to a Proof Receipt (source) Navigation, recorded proof checks and incremental validation.
From Spare Compute to Cumulative Mathematics (source) Contribution protocol, compute, credit and governance.

For current use, follow the reading guide, agent instructions or Contributing.

For the repository layout, sources of truth, build path, and release infrastructure, see docs/ARCHITECTURE.md.

Sources and earlier manuscripts

The manuscript layer (the .tex sources and rendered PDFs) is licensed CC-BY-4.0; see REUSE.toml at the repository root.

Source note for the #251 sparse construction

For #251, the results guide identifies the formal sparse construction and the differences in quantifier order between it and the printed proof. The short paper’s attribution section explains those differences before its source index.

Earlier combined manuscript: provenance only

The older joint #249/#257 paper is retained for provenance; it is not the entry point for either problem. Its arguments and release history can be inspected in the archived PDF and LaTeX source. Use the individual papers above for the current work.

Commands for agents following a paper into the source

Returning from a problem note to checked evidence

The notes are exposition, not proof authority. To return from any note to the machine-owned problem record, run the matching route below. Each problem route returns its exact paper/source record, checked module inventory, and open obligation handles; the complete eight-problem source map keeps the same joins readable. For a source-fingerprinted continuation packet, use the route-memory command in the last column.

Problem Public problem route Resumable route-memory handoff
#68 python3 scripts/query_corpus.py --route erdos_68 python3 scripts/query_route_memory.py --problem 68
#243 python3 scripts/query_corpus.py --route erdos_243 python3 scripts/query_route_memory.py --problem 243
#249 python3 scripts/query_corpus.py --route erdos_249 python3 scripts/query_route_memory.py --problem 249
#251 python3 scripts/query_corpus.py --route erdos_251 python3 scripts/query_route_memory.py --problem 251
#257 python3 scripts/query_corpus.py --route erdos_257 python3 scripts/query_route_memory.py --problem 257
#269 python3 scripts/query_corpus.py --route erdos_269 python3 scripts/query_route_memory.py --problem 269
#1041 python3 scripts/query_corpus.py --route erdos_1041 python3 scripts/query_route_memory.py --problem 1041
#1049 python3 scripts/query_corpus.py --route erdos_1049 python3 scripts/query_route_memory.py --problem 1049

Build and update

Run from the repository root, with Tectonic or a TeX Live installation:

make -C paper

The Makefile builds the manuscripts registered in docs/publication_contract.json and copies the PDFs into their problem or systems directory under paper/. The temporary build PDFs in paper/ are not the published copies.

To rebuild one paper, run make -C paper <stem>.pdf and then python3 scripts/sync_publication_pdfs.py, which copies the fresh PDF into place. Each copy is recorded in build-manifest.json with the digest of the PDF and of every TeX input it was compiled from, and a build output older than one of its inputs is refused. The release checks fail when a committed PDF is not the recorded build of its committed inputs.

The problem PDFs also depend on their generated evidence links. Tectonic keeps the .aux files so that, after a layout change, the evidence builder can read the current result numbers and pages:

python3 scripts/paper_evidence.py build --corpus-repo /path/to/plectis-erdos-lean \
  --aux-dir paper --aux-paper <paper-id>

Review and commit changed evidence records before pointing that paper’s record_commit_overrides entry in evidence/config.json at the new commit. Run the evidence builder again, rebuild the affected PDF and synchronize it. python3 scripts/check_paper_evidence_pdfs.py checks each margin link against the heading and page in the resulting PDF; make -C paper check includes this check. It needs the dependencies in scripts/requirements-release.txt.

After editing a manuscript, rebuild its PDF before updating its recorded digests. The following command previews digest changes; add --apply only after reviewing the source and rebuilt PDF:

python3 scripts/check_publication_contract.py --restamp

The statement links inside a paper identify the exact proof-source revision used for that paper. They serve reproducibility. They are not a reason to keep an older manuscript on the reading path. When the source changes, update the explanation and source links together, rebuild the PDF, and check its coverage against the current proofs:

python3 scripts/check_problem_note_sources.py --coverage
python3 docs/papers/check_paper_corpus.py

The last check rejects a Markdown mirror or recorded PDF that no longer matches its manuscript. Include the edited source, rebuilt PDF and check output in your pull request. If the remaining failure is a generated copy made stale by your manuscript edit, say so in the pull request; do not hand-edit the copy to make the check pass.

Refresh generated full text and paper-corpus records with python3 docs/papers/refresh_paper_corpus.py --write, then run python3 scripts/refresh_projections.py and python3 docs/papers/check_paper_corpus.py before merging. These owners are included in the public checkout; do not edit their generated output by hand.