Corpus and analysis
The corpus contains 153 documents and 21,422 extracted claims. Cross-document comparison grouped 6,917 claims into 2,517 groups. Theme counts use distinct document IDs.
| File | Contents |
|---|---|
| collected-texts.csv | The 2,098 collected texts listed in the collection inventory. |
| documents.csv, documents.json | Analyzed documents, source links, text hashes, and claim counts. |
| claims.jsonl | Extracted claims with source passages and document IDs. |
| stars.json | Claim groups, representatives, members, and document counts. |
| themes-topdown.json | Theme definitions and assigned groups. |
| themes-bottomup.json | Similarity clusters and their subpositions. |
| star-theme-memberships.jsonl | Group membership under each approach. |
| theme-counts.csv | Document, claim, and group counts for all 138 themes and clusters. |
| essay-claim-bullets.json | The themes, claim groups, and works behind each number the essay reports from the analysis. |
| analysis-metadata.json | Corpus and counting information. |
| expected-counts.json, recount-results.json | Expected totals and recalculated results. |
Public source URLs are recorded for 67 documents. The other records retain titles, text hashes, and source passages. A source-text hash identifies the text used for extraction; the source webpage may subsequently change.
Collected texts
The collection inventory lists 2,098 collected texts, each with a distinct document ID and source-text hash. They include 1,210 consultation responses, 303 scholarly papers, and 431 other public texts, along with smaller numbers of strategies, principles, policies, declarations, recommendations, frameworks, reports, charters, manifestos, and one standard. The inventory also includes 67 texts from 48 sources gathered for close reading (document type referenced_source), some divided into parts for extraction. These are the author’s counts of collected texts.
All 153 analyzed documents were selected from this inventory. In documents.json, each analyzed document matches its inventory entry by id or, for documents renamed during analysis, by sourceDocumentId. The inventory is not a list of documents awaiting analysis.
Each entry records the title, document type, document ID, and source-text hash. The inventory contains no full texts or dates. For 153 entries the collected text supplied no title, and some other titles are file names. Distinct records do not establish independent authors, constituencies, or endorsements, and not every collected text is a vision document.
The inventory combines 2,031 texts exported on 5 September 2026 from a frozen snapshot of the collection with the 67 texts from sources gathered for close reading. For the 2,031 exported texts, the recorded hash is that of the collected text; for the 67 added texts, it is that of the text used for extraction. The inventory file’s SHA-256 is ff97dab3960bb25f38c6efe28cb7769f1600fa98c09b26163c60ddd891cff50a.
Representative claim groups
The table gives one of the largest retained claim groups within each theme. Its representative sentence is an extracted paraphrase, not a verbatim quotation shared by the listed documents. The document count records membership in the saved group.
| Theme | Representative extracted claim | Documents in the group |
|---|---|---|
| Openness serves better science | Open Science increases trust in scientific claims. | 9 |
| The scientific record should become linked and machine-actionable | The FAIR factors of findability, accessibility, interoperability, and reuse should be strengthened for any shared research output. | 7 |
| Research should be available without delay | Everyone should be able to access the knowledge that shapes our world. | 10 |
| Research should be open to reuse | Scholarly publications should carry full reuse rights, such as those provided by a CC BY 4.0 International License or its functional equivalent. | 10 |
| Sharing research should not depend on the author’s ability to pay | Institutional repositories run by libraries or research institutions can play an important role in public access compliance because they generally do not charge authors to deposit articles or manuscripts. | 8 |
| Research should remain identifiable and connected across systems | (Meta)data should meet domain-relevant community standards. | 6 |
| Research assessment should recognize the full contribution to science | Research assessment must include greater recognition for open science practices. | 12 |
| Shared infrastructure should remain accountable to the research community | Day-to-day operations of infrastructure organisations should be supported by day-to-day sustainable revenue sources. | 10 |
| Shared infrastructure and distributed control | For an infrastructure to be forked, it is necessary to replicate all relevant data. | 7 |
| Which publishing costs institutions should pay | Article processing charges (APCs) create significant barriers for less-well-resourced researchers and institutions to make their research available. | 11 |
| Who may authorize access and reuse | The data underlying publications should be made open by default with appropriate safeguards to protect privacy. | 8 |
| Common standards and different research contexts | Any research-supporting infrastructure or service should ensure inclusivity. | 9 |
| Commercial participation and community ownership | Publishing models should be favored that are controlled by academic-led and nonprofit organizations. | 4 |
The immediate-access theme includes broader access positions and “as soon as possible” wording. Neither its 56-document total nor the ten-document group above counts explicit endorsements of zero embargo.
Related filings account for substantial membership in several groups. SPARC agency responses account for seven of the ten documents in the access group, seven of eight in the repository group, and seven of eleven in the APC-barriers group. The infrastructure-revenue group includes multiple Principles of Open Scholarly Infrastructure texts, adoption and evaluation documents, and the 2025 State of Open Infrastructure. All seven documents in the forkability group concern those principles or their adoption and evaluation. The essay’s bullets leave out groups like these: they list only groups whose documents come from several different works, as the methods explain, and essay-claim-bullets.json records each one.