The Integrated Explicit Analytic Number Theory Network: a graph of claims about explicit analytic number theory, together with a record of how well each one is justified.

Explicit analytic number theory is a field of long dependency chains. A bound on π(x) rests on a bound on ψ(x), which rests on a zero-free region, which rests on a numerical verification of the Riemann hypothesis to some height — and in the published literature those dependencies are recorded only in prose, if at all. Some links are fully rigorous, some rest on very large computations, and some rest on claims a paper states without proof.

This repository records that structure explicitly. Each node makes a mathematical claim, declares what it assumes, and carries evidence for the step from its assumptions to its claim. The evidence may be a Lean proof checked by Comparator, a numerical computation, or a citation. The goal is not that everything is certain — it is that for any result you can read off exactly what it rests on and how good that evidence is.

Every node's challenge is a conditional theorem: if the results it imports hold, then its own conclusions hold. Whether the node itself is justified is metadata, not Lean.

So a node backed by a one-line citation and a node backed by a 200,000-line machine-generated proof have the same shape, and nodes import each other's statements, never each other's proofs. The Lean layer certifies only the edges of the graph; the standing of the graph as a whole is a separate, auditable question.

IEANTN/Vocabulary/               the shared language — definitions only, Mathlib-only
IEANTN/Nodes/<Family>/<version>/ Conclusions.lean, Challenge.lean, formalization.yaml,
                                 and optionally Tables.lean for a paper's bulk data
IEANTN/Bridges/<Family>/         proofs that one version's conclusions imply another's
Solutions/<Family>.<version>/    proofs — separate Lake projects, not in the core build
receipts/                        one JSON file per Lean-verified conclusion
fingerprints.json, STATE.md,     generated and committed, so a change of meaning is a diff line
GRAPH.md, docs/nodes/
docs/                            ARCHITECTURE.md, NODES.md, ROADMAP.md, SOURCES.md,
                                 SOLUTIONS.md

The core build is Vocabulary and Nodes only, and is deliberately fast. Solution verification is a separate, dispatchable workflow: running Comparator on every pull request would take hours and defeat the purpose of the split.

0. GRAPH.md — the network itself: what it contains, what each result rests on, and what is taken on trust. Generated, and the place to start if you only want to see the thing. Each node also has a generated summary page under docs/nodes/: what it claims, in Lean and in prose, and everything recorded about why it should be believed.
1. CONTRIBUTING.md — the ten things people actually do, and how.
2. docs/ARCHITECTURE.md — the design and why it is what it is.
3. docs/NODES.md — node file formats, in detail.
4. CLAUDE.md — the short version, for coding agents.
5. docs/ROADMAP.md — what is deliberately not built yet, and why.
6. docs/SOURCES.md — the papers the nodes rest on, held and wanted.
7. SECURITY.md — the trust model, and what a forged receipt would take.

Early, but the machinery is in place and exercised. Vocabulary, three proof-of-concept node versions and the bridge between two of them, challenge generation, statement fingerprints, verification receipts, the breaking-change detector, the network checks and the housekeeping queue all exist and run in CI.

The Comparator path works end to end: Lcm.v1 carries a receipt written by the verification workflow, and ieantn.py status grades it against the current environment.

Outstanding: visualisation, the /verify comment trigger, and the composing half of the Palomar spin-off generator. docs/ROADMAP.md tracks what is deliberately not built and why; SECURITY.md states the trust model.

Open work is tracked on the IEANTN project board. Anything marked Unclaimed is free to take: comment claim on the issue and a bot assigns it to you and moves the card. disclaim releases it again, and propose PR #123 links your pull request. Claims do not expire, so please release one you are no longer working on.

This work grows out of the IEANTN subproject of PNT+, whose blueprint-based structure it is intended to replace; discussion is on the #PrimeNumberTheorem+ Zulip channel.

lake exe cache get
lake build

One declaration uses 'sorry' warning per node conclusion is expected. A challenge states; it does not prove.