What is Inferpedia?

Most encyclopedias describe what the record says. Inferpedia describes what the record implies — the things that almost certainly existed but that no surviving document directly records. The evidence of the pre-modern world is full of holes, and many of the holes have a definite shape: a letter survives that answers a letter we no longer have; a chronicle quotes a history that later vanished; a manuscript exists, so somebody — unnamed, undescribed — must have made it. Each article here is about one of those missing things, argued from the shape of the hole.

That kind of argument is easy to do badly, so the whole site is built around one discipline: every claim states how solid it is, and nothing is allowed to look more solid than it is.

The ladder

Every object on the site sits on a six-rung ladder, and climbs it in public.

• L0 — Rough source. A raw signal from the corpus: a catalogue line, a quotation, an anomaly. Noticed, not yet argued.
• L1 — Lead. Someone — a person or a model — says there is probably a missing thing here. A pointer, worth pursuing, proof of nothing.
• L2 — Candidate. The lead has survived first checks: the missing thing can be delimited, dated, and hasn't turned out to survive after all.
• L3 — Evidence packet. The dossier: every observation that implies the missing thing, each one cited and dated. This is where evidence begins.
• L4 — Draft article. The packet written up, claims graded, reviewed.
• L5 — Published article. In the encyclopedia, scores and audit trail attached.

A worked example. The Ranworth Antiphoner is a fifteenth-century English service book that still exists. At L0 it is a catalogue row noting decoration by several professional hands. At L1, a lead: a production team stood behind this book, and no document names them. At L2, a candidate: the hands are counted and characterised, and the team matches no recorded workshop. At L3, an evidence packet: the palaeographic and art-historical observations assembled, each traceable to its source. At L4, a draft about the team — how many people, what division of labour the book itself proves. At L5, the published article you can read today: not about the manuscript, but about the invisible people the manuscript could not exist without.

The evidence boundary

The line between L2 and L3 is the evidence boundary. Above it, every statement traces to a cited source. Below it live signals, leads, and open questions — shown openly, but visibly discounted: they never feed an article's scores, and nothing below the boundary is ever cited as support for anything above it.

The One Thousand and One Conjectures live below that boundary by design. They are the question-book: falsifiable conjectures about the pre-print world, each expecting to be killed, survived, or absorbed as a lead. On topic pages they appear as labelled open questions — never as evidence, and nothing on this site treats them as such.

What you can do here

Read the published articles — over 150 at preview — and follow any topic through the atlas to see what is known, what is claimed, and what is merely asked. If an article is wrong, the corrections page shows how we handle that in the open. And you can work on the question-book: if you know the study that already settles a conjecture, submit the prior art on its page; if a conjecture is malformed, dispute it; if you can run a kill-test, claim it. Verdicts change, and changes are credited by name.

Why the preview is rough

We opened the doors mid-build. Some rooms are still workshops, curated out of the navigation rather than hidden; some pages are slower than they will be; labels and layouts will shift as material climbs the ladder. What will not shift is the discipline: levels stay visible, sources stay traceable, and the untested stays labelled untested. The roughness you can see is the same roughness we see — that is the point of a research preview.