The documentation is organized by review task. Files at the top level of docs/ are maintained
mathematical or technical references. docs/orchestration/ is a historical archive of the agent
workflow that produced the proof; it is useful for provenance and design rationale, but its ticket
states are not current project status. Two files in that directory are the exception and are
live: orchestration/roe-tickets.md and
orchestration/roe-verification-plan.md, the board and
plan for the in-flight Γ_R campaign.
The original submitted paper is retained at
paper/A_Profinite_Presentation_for_G__Q_2.pdf.
The maintained writeup is published at https://roed314.github.io/gq2/. Because that writeup may
be reorganized, documentation should prefer semantic paper identifiers and Lean declaration names
over displayed theorem numbers.
| File | Purpose |
|---|---|
atlas.md |
Explains Lean Atlas, the Lean Compass semantic-review cone, the separate kernel axiom check, and how to regenerate atlas-audit.md. |
paper-api.md |
Defines the stable paper-to-Lean API policy using semantic HTML ids and data-lean-ref declarations. |
literature-axioms-onepage.md |
One-page expert checklist for the nine external mathematical inputs. |
literature-axioms.md |
Full statements, citations, encoding choices, and verification discussion for the literature inputs. |
adversarial-axioms-review.md |
Dated adversarial review of whether the Lean axiom interfaces are supported by the cited sources. Later census decisions are called out explicitly rather than silently rewriting the historical findings. |
angdinata-review-plan.md |
Response plan for an external four-point review of the axiom interfaces: the verdict on each point, the state of cup products in Mathlib and FLT, and the resulting work items. |
foundations-audit.md |
Dated survey of which required foundations existed in the pinned Mathlib revision and which gaps had to be filled locally. |
mathlib-cft-survey.md |
Dated inventory of relevant Mathlib and ClassFieldTheory APIs. It is a dependency-snapshot document, not a claim about newer upstream versions. |
cts-cohomology-gap.md |
Describes the homogeneous/inhomogeneous continuous-cohomology interface gap and the explicit low-degree model used by this project. |
| File | Purpose |
|---|---|
paper-errata.md |
Consolidated corrections, implicit hypotheses, fragile passages, and positive confirmations discovered by formalization. This is the main input for revising the paper. |
erratum-h0-transcription.md |
Detailed forensic record of the dropped d₀ factor in the first Lean transcription and its completed repair. |
off-path-statements.md |
Paper results proved in Lean but not used by the final capstone dependency route. |
section3-extraction.md |
Paper §3 statement crosswalk, absorptions, and resolved encoding deviations. |
section67-extraction.md |
Paper §§6–7 statement/display crosswalk and the mathematical amendments exposed by proof. |
section8-extraction.md |
Paper §8 crosswalk and the corrections needed for the affine-lifting and recursion argument. |
section9-extraction.md |
Paper §9 crosswalk and architecture of the terminal, block, and induction lanes. |
section10-extraction.md |
Paper §10 crosswalk for tame-frame exhaustion, equation (154), and the count-form capstone. |
tickets.md |
Concise proof-development history retained at its established path; it summarizes the mathematical route without acting as a live ticket board. |
A follow-on campaign, not part of the paper, verifies a second 4-generator 2-relator
candidate presentation Γ_R against the same finite-target machinery. Its terminal theorem
GQ2.main_presentation_literal_roe_unconditional is sorry-free, adds no axiom, and (since
2026-07-26) carries no hypothesis: the one explicit binder it used to have, BLabHypothesis, is
now the in-repo theorem GQ2.Roe.Labute.bLab, proved at the standard three axioms — see the "B3
addendum" of literature-axioms.md for the statement, the proof chain,
and the decision record.
| File | Purpose |
|---|---|
roe-campaign-summary.md |
Start here. The mathematical story in two pages: what Γ_R is, why the explicit-identification route is impossible, the route actually taken, the SourceData interface refactor, and exactly what the capstone rests on. |
orchestration/roe-verification-plan.md |
Live. The campaign plan, preceded by a dated status block recording the outcome against it — routes taken, actuals against the estimate, and the state of the follow-on Labute campaign. |
orchestration/roe-tickets.md |
Live. The ticket board: every ticket with its owned files, dependencies, landing commit, and the discoveries that changed later tickets. The provenance record for the whole campaign. |
roe-verso-chunk.md |
Prepared-not-applied blueprint chunk for the project site: the Verso markup for the Γ_R theorems in the site's own conventions, plus integration and build instructions. Nothing in it has been deployed. |
orchestration/roe-r2-spike.md |
Why no explicit word identification of the two pro-2 quotients exists — an impossibility theorem, with the independent re-derivation of the note's orientation numerics. |
orchestration/roe-r7-design.md |
The frozen statements of the pro-2 identification lane, and the first draft of the classification interface later declined as an axiom. |
orchestration/roe-r13b-plan.md |
Route for the degree-2 word-cohomology bridge that makes D_R Demushkin, including the machine-checked cup–Bockstein matrix. |
orchestration/roe-r20-recon.md |
Where the §5 duality scaffolding is generic in the relator and where it is not; the evidence behind the decision to clone the dévissage tree rather than generalize frozen code. |
orchestration/roe-r30-recon.md |
How the SourceData field list was read off the proof-time uses of Theorem 4.2, with the refactor's regression gates. |
orchestration/roe-r31-survey-ii34.md |
Supply survey: lift multiplicity, the half-torsor, and the shared obstruction layer. |
orchestration/roe-r31-survey-stage-ii5.md |
Supply survey: the display-(136) stage obligation and its cover-lift kernel. |
orchestration/roe-r31-survey-ii6.md |
Supply survey: the display-(140) residues, and the private helpers a second source has to restate. |
orchestration/roe-r31-survey-gaussz.md |
Supply survey: the Gauss-Z layer, plus the specification of the shared word substrate that unblocked the rest. |
orchestration/labute-plan.md |
Comparison of the available proof routes for the classification instance, with a page-verified literature survey and the recommended levelwise route. |
orchestration/labute-spike.md |
The computational de-risking evidence: what the naive argument gets wrong, the congruence invariant that repairs it, and the control experiment showing the invariant is necessary. |
orchestration/labute-l1-design.md |
Statement freeze for the Lean skeleton, with per-ticket proof sketches and the deviations taken against the spike. |
Several important review artifacts live at repository root because external tools expect them there:
../formalization.yaml— structured provenance, fidelity, automation, alignment, and review metadata;../comparator-config.jsonwith../Challenge.leanand../Solution.lean— the Comparator validation pair;../atlas-audit.md— committed, regenerated Lean Compass and kernel-axiom report;../scripts/check_axioms.sh— the primary gate: axiom census,sorry/native_decidehygiene, and the per-capstone expected-axiom audit;../scripts/atlas_audit.pyand../scripts/paper_api_audit.py— report generators;../scripts/roe_sanity_counts.py— independent finite-quotient count check for theΓ_Radmissible-marking semantics.
orchestration/ preserves the plans, handoffs, counterexamples, and design
records produced during the July 2026 agent-assisted formalization. These files often contain
valuable mathematical reasoning, but words such as “open”, “blocked”, “remaining”, and “sorried”
describe the moment when the note was written. They must not be used as a current ledger — with
the two exceptions named above, roe-tickets.md and roe-verification-plan.md, which are the
live board and plan of the Γ_R campaign.
The archive's README.md lists every retained file and explains its
historical role. Current truth comes from the Lean sources and mechanical gates:
lake build
bash scripts/check_axioms.sh
lake env lean GQ2/AxiomLedger.lean
python3 scripts/atlas_audit.py atlas-graph.json