Limited published evidence · History of mathematics and research documentation
Reported Gödel-notebook digest and candidate-proof review
A secondary Substack roundup points to a Guardian interview about research begun in September. The digest and proof-checking claims are attributed and independently unverified. The proposed rehearsal below is separate and unexecuted.
Review a digest and candidate proofs from Gödel’s notebooks
Jan von Plato’s Helsinki Gödel-notebook research project
Interview-reported research begun in September, covered by The Guardian on 13 November 2026 and a secondary Substack roundup on 14 November; mathematical results not independently verified.
A published third-party example, not a Nuvepro customer case or evidence of a partnership.
The workflow at a glance
Reported notebook analysis combines AI drafts and human review
Illustrative summary of the published account
Reported work components. Their arrangement does not imply a fixed execution order or simultaneous work.
- Jan von Plato and Helsinki project
Supply transcriptions and contextual literature
The interview describes giving Claude existing human transcriptions, von Plato’s books and related literature.
- Claude
Produce a digest and candidate statements
The reporting describes an AI digest and extraction of statements with proofs, sketches or ideas for proofs.
- Historian
Review the digest page by page
Von Plato reports checking the digest page by page. The source does not establish independent acceptance or publication authority.
- Researcher and reported Coq checks
Check candidate proofs and distinguish repairs
Von Plato reports Coq checks, statements needing more precise definitions or assumptions, and false results. Proof files and checking runs were not inspected.
What the source reports
The Guardian reports that Jan von Plato supplied existing human transcriptions of Gödel’s historical notebooks, his books and related literature to Claude. The model produced a digest and candidate mathematical statements; the historian describes page-by-page review. He reports Coq checking of reconstructed results, with some statements repaired through clearer definitions or added assumptions and others found false. This is an attributed research account, not inspected proof evidence.
Reported technology: Claude analysis with historian review and reported Coq checking. Scope and outcomes are source-reported.
What the enterprise environment would need
The following is a proposed project design for an enterprise doing similar work. Published details are marked; the remaining systems, data and scenarios would be agreed for your project.
Proposed practice environment
Review a source-linked digest and proof rehearsal
Illustrative project design to scope with your enterprise.
- 1 · Set up the environment
Permitted transcriptions and frozen references
Use approved or synthetic transcription passages, contextual material, rights records and independently authored expected statement variants. Freeze source versions, notation, definitions, assumptions and a proof-checker version/configuration in an isolated rehearsal.
- 2 · Preparation and agent contributions
Agent drafts linked statements and candidate proofs
Prepare a digest, passage citations, candidate statements, explicit assumptions and proof attempts. Keep missing text and ambiguous notation unresolved; preserve every change and generate reviewable artifacts only.
- 3 · Independent checks
Separate mathematical and source-fidelity checks
A mathematician separate from the generating agent compares statements, quantifiers, definitions and assumptions with the approved passages and reference variants. Independently run the frozen formal checker against known valid/invalid reference cases and candidate proof files. A checker pass cannot verify historical meaning or transcription fidelity.
- 4 · People decide acceptance
Historian accepts the exact rehearsal package
The accountable historian accepts or rejects the exact independently checked digest, statement/proof versions, citations and findings together. Acceptance is limited to this rehearsal and grants no original-proof, attribution or publication authority.
A reviewable evidence package
A versioned rehearsal digest, source-to-statement map, proof attempts, checker records, independent semantic findings and owner disposition.
If a check fails → withhold acceptance.
Missing permission, uncertain transcription, changed assumptions, failed formal checks, a source mismatch or owner rejection withholds acceptance. Repair the affected digest, statement and proof package, repeat independent source/semantic and formal checks, then return the exact revised version to the historian.
Failure or owner rejection returns to review
- Hold
- → 2 · Revise work
- → 3 · Rerun independent checks
- → 4 · Owner review
Operating boundary. Designed checks, not executed. Approved or synthetic rehearsal only; no real archive upload, publication, theorem-validity claim or human-readiness proof. Mathematical, historical and rights owners must confirm references and acceptance criteria.
- Permitted transcriptions and frozen mathematical references
- Proposed: a research owner provides licensed or synthetic transcriptions, source identifiers, permitted contextual literature and specialist-prepared statement variants with explicit definitions and assumptions. Freeze the proof-checker version, dependencies and expected results; do not use real archive material without permission.
- Source-linked digest and candidate proof drafts
- Proposed: an agent drafts a digest, candidate statements and attempted formalizations in an isolated workspace. Keep source attribution, inferred additions, unsupported steps and unresolved statements separate. An agent cannot declare a historical discovery or publish a proof.
- Independent mathematical and historical checks
- Proposed: a mathematical specialist separate from the generating agent compares each statement and formalization with the frozen reference, runs the frozen proof checker and checks for changed assumptions or admitted gaps. A separate source review checks transcription fidelity and distinguishes manuscript content from modern reconstruction; checker success alone does not establish historical attribution.
- Research-owner acceptance and repair
- Proposed: the accountable research owner accepts or rejects the exact independently checked rehearsal digest, statement, formalization and checking record. False statements, missing assumptions, changed inputs or checker errors block acceptance. Repair the affected package, repeat independent mathematical and source checks and return the new version to the owner.
A proposed simulation
An illustrative practice design derived from this work, not an executed assessment.
Proposed: prepare a source-linked rehearsal digest and candidate-proof package from permitted or synthetic transcriptions, retaining explicit assumptions, unresolved claims and independent checking records.
Review gate: The research owner accepts only the exact rehearsal package whose source fidelity, statements, assumptions and formal-checking records passed independent specialist checks. Checker success does not establish historical attribution or authorize publication.
Challenge the result: A candidate passes only after an unstated assumption is added. Withhold acceptance, expose the change, repair the statement and attribution, repeat independent mathematical and source checks and return the exact revised package to the research owner.
What needs checking
- Source boundary: Shared Sapience is a secondary roundup and The Guardian is interview reporting. Neither supplies inspected proofs or independent mathematical validation; the external logician’s assessment was conditional and he had not seen the reconstruction.
- Proposed: test a supported statement, a missing assumption, a false candidate, a transcription error, an admitted proof gap and a checker/dependency mismatch against specialist-prepared references. Keep checker validity, source fidelity and historical attribution as distinct decisions. Designed, not executed.
- Proposed: alter a definition or source page after review. Withdraw acceptance, repair the digest, statement and affected formalization, repeat independent checking in the frozen environment and return the exact revised package to the research owner. Designed, not executed.
Review environment fidelity, human capability and agent performance separately. The enterprise’s accountable owners decide acceptance of the shared workflow and any operating use.
How an enterprise could build similar work
A proposed build order for an enterprise doing similar work, written by Nuvepro from the published account and our interpretation of the environment and checks above. It is not how Jan von Plato’s Helsinki Gödel-notebook research project built or operates this work.
- Freeze permitted sources and formal referencesProposed: agree licensed or synthetic transcriptions, statement variants, assumptions, specialist expected results and proof-checker dependencies.
- Draft source-linked research materialProposed: prepare a digest, candidate statements and formalizations while marking inferred additions and unresolved proof steps.
- Check mathematics and attribution independentlyProposed: separate specialists compare source fidelity and formalized statements with frozen references and run the frozen checker; keep historical attribution distinct.
- Repair and repeat independent checksProposed: block false statements, missing assumptions and checker errors; repair all affected material and recheck the revised version.
- Research owner accepts the exact rehearsal packageProposed: accept or reject the checked digest, statements, formalizations and records without authorizing publication.
Your systems, data, rules and accountable owners decide the real project. Agree acceptance criteria, escalation owners and operating responsibility with the people accountable for the work before anything runs on live records. Treat the first run as a rehearsal on test records: compare the draft output with an independently prepared answer, and record what the people reviewing it change.
Questions about this example
- Did this curation verify Gödel’s reconstructed proofs?
- No. The source is interview reporting, not released proof artifacts or peer review. Raw transcriptions, prompts, digest, proof files and Coq environment were not inspected or run.
- What did people and AI reportedly do?
- Human transcriptions, authored books and literature supplied the input; Claude produced a digest and candidate statements. This is not evidence of autonomous shorthand deciphering from raw manuscript images. The historian reports page-by-page review and Coq checking, with repaired and false statements distinguished.
- Is November coverage evidence of a new deployment?
- No. The Guardian interview is dated 13 November 2026 at 16:38 GMT and the Substack roundup 14 November; the work reportedly began in September and concerns historical notebooks.
- What would the proposed rehearsal establish?
- Independent specialists would check source fidelity, explicit assumptions and formal-checker results against frozen references before exact-version research-owner acceptance. These checks are designed, not executed; checker success does not establish historical attribution or publication authority.
Shape this around your enterprise
Bring a similar workflow and the people responsible for it. Together, define the records and systems to use, the result your team will build or configure, and how its owners will check it. Nuvepro scopes the simulation, environment and preparation. Agree any existing-agent evaluation, implementation assistance and production responsibilities separately.
Sources and limitations
Curation review: . This is the date this source example was reviewed, separate from source publication dates.
- The public written source captures were read. Raw transcriptions, books, model version, prompts, digest, proof files and Coq checking environment were not inspected or run; no proof validity or numerical gain is independently verified.
- Claude received existing human transcriptions and contextual literature. This account does not establish autonomous deciphering of raw manuscript images or Gödel’s shorthand. Reported formal-checker success alone does not establish semantic fidelity to the source.
- The Guardian interview was published 13 November 2026 at 16:38 GMT; the secondary Substack coverage is dated 14 November. The work reportedly began in September and concerns historical notebooks, not a new November deployment.
- The quoted external logician had not inspected the reconstruction and offered a conditional assessment. Historical attribution remains a separate interpretive question; source-reported proof checking is not independent peer review.
- Archive permissions, the complete repair procedure and final research-publication authority are not established. No FDE delivery, Nuvepro customer or partnership is claimed.
- The separate proposed rehearsal and business checks are designed and unexecuted. Mathematical and archival owners must confirm references and acceptance rules; rehearsal acceptance grants no research-publication authority or learner-readiness claim.
- Shared Sapience: Claude unlocks Gödel’s notebooks as research gates give waySource publication: 2026-11-14
- The Guardian: AI helps unlock discoveries hidden in Gödel’s notebooksSource publication: 2026-11-13
More research and regulated documentation examples
- RA3I’s described photograph-to-tracing documentation workflowRA3I platform described by Seth Chagi
- AI analytics failure investigation: Bayer Crop Science exampleBayer Crop Science
- AI batch record review: Enveda exampleEnveda
- AI clinical policy digitization: Cohere Health exampleCohere Health