Registry v1.0.0 · Updated 2026-07-27

Maha Research Context Registry

A public index of versioned, source-aware research artifacts designed for human reading and bounded AI-assisted use.

This registry is not a journal, a peer-review system, or a scientific authority. It indexes artifacts and records what each one is, what it is based on, and what it does not establish. Listing an artifact here says nothing about whether the field agrees with it.

How to read this registry

What a research context pack is

A context pack is a bounded, versioned package of a research artifact: its claims, each with a stable identifier and an epistemic status; its sources, each traced to where it was verified; and its exclusions, stated as part of the package rather than left to inference. The point is that the qualifications travel with the content. A claim separated from its status label and its limitations has lost the thing that made it safe to reuse — and that separation is exactly what happens when research is scraped, summarized, or pasted into a model without provenance.

Why claims, sources, versioning, and exclusions matter

A claim identifier lets someone cite a specific formulation rather than a paraphrase. A source trail lets them check it. A version and review date let them tell whether the copy they hold is current. An exclusion tells them what the artifact was never meant to support. Together these are what separate a reusable research artifact from a confident-sounding document — and they are what a downstream reader, human or machine, needs in order to represent the work honestly.

Four different things, deliberately not blurred

Research atlas
An orientation layer over a body of literature. It maps how concepts connect and labels how well each claim is supported. It is not a source; it points at sources.
Context pack
A bounded, versioned package of an atlas or dataset — claims, sources, exclusions, and citation metadata — designed to be reused, including as context for an AI-assisted workflow, without losing its provenance.
Working paper
A self-published research document with a stated status and version. Not peer reviewed unless it explicitly says so, and never to be represented as a journal article.
Dataset
Structured data with documented provenance and reuse terms. No dataset is currently registered.
Methodology or standard
A document specifying how research artifacts are produced, versioned, sourced, or audited. It carries no scientific findings of its own, and self-published methodology is not an accredited standard.
Peer-reviewed publication
A paper that has passed independent expert review organized by a journal or conference. Nothing in this registry is peer reviewed. No artifact here should be cited or described as though it were.

Inclusion in this registry does not imply scientific consensus. It records that an artifact’s provenance, status, and boundaries have been made explicit and can be checked — nothing more. Several registered artifacts describe debates the field has not settled, and they say so.

The rules those properties are held to are written down in the Maha Provenance Standard, which is itself registered here. It is self-published and self-assessed, and each of its clauses links to the artifact where the practice can be checked.

Registered artifacts — 4

An artifact is listed only when its canonical URL, machine-readable endpoints, status, and any DOI can be verified against the source repository. Machine-readable form at registry.json.

Filter the registry

Artifact type
Topic
Availability

Showing 4 of 4 artifacts

  • Research atlasv1.2.0DOI

    The de Sitter / String Swampland Atlas

    Public educational orientation tool

    Not peer reviewed. Does not adjudicate whether string theory admits metastable de Sitter vacua, and asserts no experimental confirmation of string theory, M-theory, or any swampland conjecture.

    A source-first concept map and claim ledger for the de Sitter problem in the string/M-theory swampland program. Sixteen concept nodes, twelve claims with stable identifiers and epistemic status labels, seven open problems presented with both camps, and twenty-eight annotated source cards — every citation drawn from the verified set of the underlying working paper. Ships with a downloadable context pack carrying its own exclusions.

    • Claim ledger (12)
    • Source trail (28)
    • Machine-readable JSON
    • Downloadable context pack
    • Stable claim URLs
    Machine-readable (7)
    • Atlas metadata · application/jsonConcept nodes, edges, and the claim ledger.
    • Claim ledger · application/jsonClaims with status, limitations, source ids, review date, canonical URL.
    • Source trail · application/jsonSource cards with verification labels and provenance notes.
    • Context pack manifest · application/jsonIntended uses, exclusions, endpoints, integrity and citation notes.
    • Plain-text context · text/plainCompact context document for AI-assisted workflows.
    • Context pack download · application/jsonSingle-file bundle: manifest, claims, sources, open questions, README.
    • Citation metadata · text/yaml
    Archive
    DOI 10.5281/zenodo.21603961 ↗ · concept DOI 10.5281/zenodo.21603960 ↗This DOI identifies the underlying working paper, not the atlas. The atlas and its context pack have no DOI of their own.
    Intended use
    • Students and self-directed learners: Orient yourself in a contested literature before reading the primary papers, with an explicit signal of which statements are settled and which are not.
    • Educators: A worked example for teaching the difference between an established result, an open research question, a conjecture, and a speculative interpretation — using a live debate rather than a tidy one.
    • Researchers entering the area: A provenance-checked starting bibliography of twenty-seven works, with the structure of the seven open problems and the strongest argument on each side.
    • AI-assisted research workflows: A compact, bounded context attachment that carries source links and epistemic labels with it, so a model summarizing this material has the qualifications in front of it rather than having to infer them.
    What this does not establish
    • This pack is not peer reviewed and confers no authority. It is an educational orientation resource assembled by a non-specialist curator.
    • It is not a complete string theory corpus. It covers one debate — whether string/M-theory admits metastable de Sitter vacua — through twenty-seven works, and nothing else.
    • It asserts no experimental confirmation of string theory, M-theory, or any swampland conjecture. None of these has been experimentally established, and the pack must not be used to suggest otherwise.
    • It does not adjudicate the central question. Competent researchers hold opposing positions for technical reasons, and the pack represents both without deciding between them.
    • It contains no full text, abstracts, figures, or data from any cited paper. Sources are identified and linked, never reproduced.
    • It is not a substitute for the primary literature. Research conclusions require reading the cited papers.
    • Its citation set is deliberately closed to the source paper. Works outside that set — including the discovery-era supernova cosmology papers — are absent by design, and gaps are stated rather than filled.
    • It contains no numerical results, datasets, or code, and makes no predictions of its own.

    Updated 2026-07-27 · Reviewed 2026-07-27 · CC BY 4.0

    String theoryQuantum gravityCosmology

  • Research atlasv1.1.0

    Synthetic Intelligence Atlas

    Public educational orientation tool

    Not peer reviewed. Reports no benchmark scores and makes no capability forecast. Evidence cutoff 2026-06-14; nothing here asserts a fact after that date.

    A source-bounded map of large language models, agents, evaluation, and the limits of capability forecasting. Every public claim carries content-verified sources, an epistemic status, and a limitations statement. The map makes structural claims about what benchmarks measure rather than reproducing leaderboard scores, and separates model capability from scaffolding capability throughout.

    • Claim ledger (22)
    • Concept registry (18)
    • Source trail (33)
    • Machine-readable JSON
    • Stable claim URLs
    • No DOI
    Machine-readable (4)
    • Atlas metadata · application/jsonConcepts, claims, sources, and the scope statement.
    • Claim ledger · application/jsonPublic claims with status, limitations, concept ids, source ids, and review date.
    • Source trail · application/jsonContent-verified sources with identifier, canonical URL, type, and why each is cited.
    • Plain-text context · text/plainCompact context document for AI-assisted workflows.
    Intended use
    • Readers orienting themselves in the literature on language models, agents, and evaluation, who need to know which statements are settled and which are contested.
    • Educators teaching the difference between a benchmark score and a capability claim.
    • AI-assisted research workflows that need bounded context carrying its own sources, status labels, and limitations.
    What this does not establish
    • Not peer reviewed and not an authority. It is an educational orientation layer assembled by a non-specialist curator.
    • Reports no benchmark scores, leaderboard positions, model rankings, price trajectories, or compute estimates. It makes structural claims about what benchmarks measure instead.
    • Makes no capability forecast. Conditional scenarios are held in the audit record and are not part of this public edition.
    • Quantum computing is out of scope by design and requires a separate atlas with different primary literature.
    • Provider technical reports and model cards are recorded as self-reports, never as independent measurements.
    • Evidence cutoff is 2026-06-14. Nothing in the map asserts a fact after that date, and sources were resolved on 2026-07-27.
    • It has no DOI. Cite it by URL, atlas version, and claim identifier.

    Updated 2026-07-28 · Reviewed 2026-07-28 · CC BY 4.0

    Artificial intelligenceMachine learningEvaluation

  • Working paperv2.0DOI

    The de Sitter Problem in the String Swampland: A Verified Literature Map

    Preprint · Not peer reviewed

    A literature map, not original research. It surveys a contested debate and represents the disagreement rather than resolving it.

    A survey of the de Sitter problem in string/M-theory structured around seven open problems and their competing camps. Every cited arXiv identifier was independently resolved against arXiv and INSPIRE-HEP before publication, and provenance tags record the verification status of each cited work. This paper is the authority for every citation used by the atlas built on it.

    • Provenance-tagged citations
    • Verification ledger
    • Machine-readable metadata
    • Archived with DOI
    Machine-readable (4)
    Archive
    DOI 10.5281/zenodo.21603961 ↗ · concept DOI 10.5281/zenodo.21603960 ↗This DOI identifies this paper. Cite the version DOI for the archived edition you consulted.
    Intended use
    • Researchers and students who need the structure of the de Sitter debate and a provenance-checked entry point into its literature.
    • Anyone assessing what an AI-assisted synthesis can and cannot be trusted to produce: the paper documents its own verification passes and corrections.
    What this does not establish
    • Not peer reviewed, and must not be represented as a journal article.
    • Not original research. It maps a debate; it advances no physics position of its own.
    • It does not adjudicate whether string theory admits metastable de Sitter vacua, and takes no side on the question.
    • Verification means a citation resolves to a real paper correctly placed in the debate. It is not an endorsement of that paper’s argument — most works cited disagree with one another.
    • Affiliations and camp placements are time-stamped to the revision date, not permanent.

    Updated 2026-07-26 · Reviewed 2026-07-26 · CC BY 4.0

    String theoryQuantum gravityCosmology

  • Methodology or standardv1.0.0

    The Maha Provenance Standard

    Internal methodology specification

    Not a peer-reviewed standard, not issued by a standards body, and not adopted outside this project. It documents how one research project works, so that its outputs can be checked against a stated rule rather than taken on trust.

    A claim-level provenance framework specifying how an assertion is given a stable identifier, an epistemic status, a traceable source, and an explicit statement of what it does not establish — and how those properties survive packaging, export, and reuse by an AI system. Fourteen clauses across provenance, claim structure, separation of judgement from fact, versioning, and machine reuse, each linking to where it is enforced. Descriptive of practice already in use, not proposed for adoption.

    • Clauses (14)
    • Evidence links per clause
    • Conformance levels
    • Machine-readable JSON
    • No DOI
    Machine-readable (1)
    • Standard record · application/jsonClauses with rules, rationale, and evidence URLs; conformance levels and self-assessed conformance.
    Intended use
    • Readers assessing how far a Maha artifact can be trusted, who want the rule stated rather than inferred from the output.
    • Anyone building self-published or AI-assisted research who wants a worked example of attaching provenance and epistemic status at the level of the individual claim.
    What this does not establish
    • Not a peer-reviewed standard and not issued by any standards body.
    • Not adopted or in use outside Maha Strategies Research, and not proposed for anyone else to adopt.
    • Conformance is self-assessed. No external body has audited the level assignments; the evidence links exist so a reader can check them.
    • Descriptive rather than prescriptive: where a clause and the codebase disagree, the codebase is the fact and the clause is the error.
    • It is not a general research methodology and not a substitute for peer review.
    • It has no DOI. Cite it by URL and version.

    Updated 2026-07-27 · Reviewed 2026-07-27 · CC BY 4.0

    Research methodologyProvenance

Inclusion policy

  • An artifact is listed only when its canonical URL, machine-readable endpoints, status, and any DOI can be verified against the source repository.
  • Inclusion records provenance. It is not a claim that the field agrees with an artifact, and it is not peer review.
  • Each entry states what its artifact does not establish. Those exclusions travel with the record and should be preserved on reuse.
  • Artifacts whose metadata is incomplete are listed under pendingArtifacts with the specific gap named, rather than being included on the strength of a description.

Registry v1.0.0, updated 2026-07-27. Machine-readable at /registry.json. Licensed CC BY 4.0. Individual artifacts carry their own license and status.

Published by Maha Strategies under the direction of Mayone Maha Rajan. No login, no paywall, no tracking.