RESEARCH & KNOWLEDGE / WORKING PAPER INITIATIVES
Publication Accessibility, Evidence Continuity and AI Research
Three linked working manuscripts — initiated 13 November 2026. These research records connect an implemented source-linked public archive, a proposed voluntary or regulatory standard, and a proposed educational and AI evaluation programme. They are not peer-reviewed journal articles, accepted publications or completed experiments.
Open downloadable archive example · Original generator code · Implementation and reuse instructions.
Research paper initiatives
WORKING PAPER A · DESIGN SCIENCE / SOFTWARE ARCHITECTURE
Provenance-preserving period-based public publication archives
Specifies the extraction pipeline, dated record model, daily-to-annual PDF generation, machine-readable exports, public interfaces, architecture evaluation and falsifiable completeness tests. Explicitly distinguishes dated work logs from original dates of externally linked publications.
Read complete working manuscript A
WORKING PAPER B · COMPARATIVE LEGAL AND PUBLIC POLICY ANALYSIS
Public-interest publication accessibility and archival duties
Investigates the feasibility, proportionality and safeguards of permanent publication directories, regular accessible bundles, correction records and historical transparency. The suggested duty is a research proposal, not a universal obligation currently established in law.
Read complete working manuscript B
WORKING PAPER C · EDUCATION / AI EVALUATION / KNOWLEDGE CONTINUITY
Living evidence archives for higher education and advanced AI evaluation
Examines how original dated sources, accessible archives, correction histories and code-based exercises might complement textbooks, foster evidence literacy and strengthen longitudinal AI evaluation. Educational impact and long-term model reliability are hypotheses to test.
Read complete working manuscript C
Code genealogy, authorisation and reproducibility
The generator was authored as a new implementation, not cloned from another repository. It uses established, attributed tools. Code visibility does not grant rights to third-party articles, proprietary archives or copyrighted full text.
Project levels and measured outcomes
Beginner — understand and reproduce a small dated index
Read the architecture paper, inspect owned/public source log records, generate one day and one month, compare JSON against PDF links and identify the difference between a log date and a publisher's original release date. Evidence: annotated script, simple test and a source critique.
Intermediate — extend the archive and accessibility tests
Adapt an expressly permitted CMS export, validate dates and canonical URLs, detect duplicate entries, add corrections and publication statuses, audit keyboard navigation, compare HTML, JSON and PDF retrieval tasks, and document CI coverage. Evidence: modular implementation, controlled tests and reproducible benchmark.
Advanced — independent assurance and multi-source longitudinal evaluation
Use licensed or otherwise authorised publication catalogues with independent coverage audits; assess version changes, evidence contradictions, human oversight, preservation of withdrawn records without unlawful republication, cross-institution source metadata and historical educational/AI evaluation. Evidence: threat model, human-subjects ethics, statistical analysis, negative results and independent replication.
Limits, safeguards and reproducibility
This reference archive indexes Manifold Grace's public dated log, not all of its pages and not every article ever written. It does not train an AI model or validate model improvement. The generated PDFs are not independently certified to PDF/UA; alternative HTML and JSON/CSV are available for accessible and machine-oriented uses. Copyright, database rights, privacy and lawful access remain separate constraints.
Research checks should compare log entries to generated metadata, detect missing dated blocks, verify original-source links, inspect rendered PDF pages, perform independent accessibility assessment, review updated scripts and distinguish GitHub commit verification from live website publication.
Evidence and source register — original publication dates
This register distinguishes primary standards and legal guidance, peer-reviewed evidence, historic archive examples and software dependencies. The date is the original verified release/publication date where available; subsequent corrections and updates remain distinct. Living documentation is labelled without inventing a date. Source links support the stated research context, not claims of completed experimental validation or an existing universal PDF publication duty.
Implementation provenance: the archive generator is original Manifold Grace code, not a clone; the source dependencies have their own terms. Generator source · Limited generator reuse licence · Versioned workflow.