Knowledge Architecture
The Open Human Record Initiative
Custody of an archive is not ownership of the knowledge inside it. Historically significant material should be presumptively accessible to the public it belongs to.
Governments, religious institutions, museums, universities, and private collections hold enormous bodies of historical material — some of it unique or inaccessible anywhere else. Preserving those materials is a real responsibility. It should not automatically carry permanent control over what the materials reveal.
The core principles
- Inventory before secrecy. The public should be able to discover what a collection contains, whether it's catalogued or digitized, and why any of it is restricted — even before requesting the material itself.
- Digitize the record. Preservation of a fragile original and public access to its information are separate problems. Digitization should carry full provenance, not just content.
- Direct harm, not discomfort. Restriction requires a specific, demonstrable risk — not embarrassment, controversy, or conflict with accepted narratives.
- Protect people, not institutions. The least restrictive fix should always be preferred: redact three lines rather than sealing two hundred pages.
- No permanent restriction by default. Every restriction should name what was withheld, why, by whom, and when it comes up for review again.
- The custodian isn't the only judge. Contested restrictions should be reviewable by an independent body, with the burden on whoever wants continued secrecy.
- Discovery is part of access. Catalogues should let researchers explore what they don't already know exists — not just retrieve documents by name.
A technical path: human-supervised archival AI
The scale of the world's archives — transcription, translation, cataloguing, cross-referencing, provenance tracking — is a real bottleneck. AI can meaningfully reduce it, assisting with OCR correction, handwriting interpretation, translation, metadata generation, duplicate detection, and confidence reporting, while leaving interpretation of what history means to human review. Every machine-generated output stays distinguishable from, and checkable against, the original source.
Disclosure: this initiative's author, Jason McCoy, is the founder of Pre-Failure Research Group and the developer of OmniSniffer. OmniSniffer is proposed as one candidate architecture for the workflow above — named because it's the system the author has built and understands in depth, not because it has been evaluated against competing approaches. It is not a required or exclusive foundation for the initiative.
Provenance: this document was drafted collaboratively — the argument and positions are Jason McCoy's, developed and refined through discussion with two AI systems, Claude (Anthropic) and GPT (OpenAI), which assisted with structuring, editing, and stress-testing the disclosure language above. As with PFRG's other AI-assisted work, this credit reflects documented contribution to the drafting process, not authorship, ownership, or personhood.
No authority over truth
This initiative doesn't assume any particular archive is hiding something extraordinary. It assumes something narrower: humanity should be able to examine its own historical record and find out what the evidence actually says. If accepted history is correct, open records strengthen it. If it's incomplete, open records let it improve.