Scholars spend lifetimes noticing that a Sumerian flood story predates Genesis by a thousand years, or that the Buddha and Osiris share the same death-and-return arc. We built a system that finds these patterns across every tradition, simultaneously — and shows its work.
The world's sacred texts — the Quran, the Bhagavad Gita, the Book of Enoch, the Tao Te Ching, the Zohar — were written across 4,000 years, in a dozen languages, on three continents. They contain overlapping flood myths, parallel creation stories, mirrored hero journeys, and suspiciously similar cosmologies. But connecting them has always required a scholar who happened to read both Akkadian cuneiform and Sanskrit. That's vanishingly rare.
The result? The most interesting questions in comparative religion — did the dying-and-rising god motif originate once and spread, or did every culture invent it independently? — remain trapped behind institutional gates, scattered across journal articles that cite each other in circles.
We thought: what if you could ask that question to all 26 texts at once, and get an answer that traces every claim back to its source, scores its confidence, and tells you where the scholars disagree?
We don't just store text. Every passage lives in a knowledge graph — a web of relationships where Marduk is linked to Tiamat is linked to the concept of primordial chaos is linked to Genesis 1:2 ("the earth was without form, and void") is linked to the Enuma Elish's Tablet I. Every node has a type. Every edge has a confidence score. Every claim has a citation.
Alongside the graph, a vector database stores 1024-dimensional semantic embeddings of every passage. This means we can find connections that aren't lexical — passages that mean similar things even when they use completely different words, in different languages, from different millennia.
voyage-3-large model. This is the magic that lets us find
semantic siblings — passages that feel alike even across languages and millennia.
A Sumerian prayer and a Vedic hymn addressing the same cosmic question will cluster together
in vector space, even though they share zero vocabulary.
This is what separates us from "vibes-based" comparisons. When The Archive says "Gilgamesh's flood narrative likely influenced the Genesis account", it doesn't just assert it. It scores the claim across six weighted dimensions:
The AI ceiling: Any connection based solely on AI inference — without corroborating human scholarship, archaeological evidence, or linguistic proof — is capped at 0.7 confidence. The system literally cannot claim certainty about something only it has noticed. It has to show its work.
Connections scoring below 0.3, or where scholarly consensus is below 0.4, are automatically flagged as disputed. They're still visible — interesting hypotheses are valuable — but they're clearly marked so nobody mistakes an emerging pattern for established fact.
When the cross-reference analyzer compares two entities from different traditions, it doesn't just say "these are similar." It classifies the mechanism:
Every cross-reference also stores falsification criteria: what evidence would disprove the connection? And counter-evidence: what existing scholarship argues against it? This isn't a system that confirms what you want to hear. It's a system that tries to prove itself wrong.
Beyond entity extraction, we map every narrative against three foundational frameworks from comparative mythology and folklore studies:
Vladimir Propp's 31 Narrative Functions — The recurring building blocks of folk tales, from "the hero receives a magical agent" to "the villain is punished." We tag passages with their Propp function, then compare across traditions. It turns out the Sumerian Descent of Inanna and the Greek Persephone myth hit many of the same functions in nearly the same order.
Joseph Campbell's Monomyth — The 17 stages of the Hero's Journey, from "The Call to Adventure" through "The Return." We map where each sacred text's narrative falls on this arc, enabling cross-tradition comparison at the structural level.
Thompson Motif Index — The standard classification system for recurring elements in world folklore, with 22 categories from A (Mythological Motifs) to Z (Miscellaneous). When you see that motif A1010 ("World-flood") appears in Mesopotamian, Hebrew, Hindu, and Norse traditions, that's the Thompson Index at work.
Say you ask: "Is there a connection between the Egyptian weighing of the heart and the Christian Last Judgment?"
The system doesn't guess. It:
1. Finds the relevant passages via semantic search (vector similarity across the
Book of the Dead and Revelation).
2. Retrieves the extracted entities (Ma'at, Osiris, Anubis / Christ, the Lamb, the Book of Life)
and their relationships from the knowledge graph.
3. Checks the cross-reference table for existing Osiris–Christ scholarly analysis
(Frazer, Mettinger, Smith — with dates, claims, and counter-claims).
4. Scores the connection across all six confidence dimensions.
5. Synthesizes an answer in your chosen reasoning mode — authoritative, transparent,
dialectical, or Socratic — with every claim hyperlinked to its source passage.
"And I saw the dead, great and small, standing before the throne, and books were opened."
Revelation 20:12 (KJV)"O my heart of my mother! Do not stand up as a witness against me, do not be opposed to me in the tribunal."
Egyptian Book of the Dead, Spell 30B (Faulkner)Two texts, two millennia apart, two completely different cultures — and the same existential architecture: a cosmic courtroom where the dead are judged by the contents of their heart. The Archive doesn't flatten these into equivalence. It maps exactly how similar they are, why they might be connected, and who disagrees.
I built this because I kept hitting the same wall. I'd read the Bhagavad Gita and think "this sounds exactly like what the Stoics said" — and then spend three hours on JSTOR trying to find out if anyone had noticed that, only to discover it was buried in a 1987 monograph behind a $40 paywall. That's insane. These are humanity's most important texts. The connections between them shouldn't be locked up.
This isn't a toy. The confidence scoring system means you can trust the connections at face value — or drill into the evidence chain if you're skeptical. The AI ceiling means the system can't hallucinate certainty. The falsification criteria mean every hypothesis comes pre-loaded with the seeds of its own disproval.
And we're just getting started. 21 passages enriched so far, with ~1,700 more to go. Every passage we process adds entities, themes, and cross-references to the graph. The intelligence compounds. A connection that's invisible with 100 entities becomes obvious with 1,000.
The goal is simple: make the study of comparative religion as navigable as Google Maps made geography. Zoom out, see the patterns. Zoom in, read the primary sources. Every level of detail is grounded in evidence.
A tool that makes humanity's deepest questions navigable. Not by simplifying them — by finally giving them the structure they deserve.
© 2026 Moonlit Social Labs · The Archive v0.6.0