How the distillation works

The four layers

The distillation proceeds in four layers, each preserving the layer beneath it:

N=1 — Per-text atomic statements

Each scripture book or chapter is read closely and its atomic moral and metaphysical claims are extracted, each anchored to specific verses. For the Dhammapada this means principles per chapter; for the Bible, per book of the canon. These are the "cards" of the catalog.

N=2 — Within-tradition convergence

Where a tradition has multiple scriptures (Buddhism's Dhammapada + four nikāyas, Hinduism's Gītā + Upaniṣads, Confucianism's Analects + Great Learning + Doctrine of the Mean + Mencius), the N=1 statements are folded together. New principles emerge only where the additional sources add something the first didn't.

N=3 — Synthesized per-tradition principle set

A minimal operational principle set per tradition (typically 14-18 principles, deliberately thinner for Shinto at N=8). Each principle is checked for structural completeness against canonical taxonomies (e.g. Buddhism's Four Noble Truths, the Eightfold Path, the Brahmavihāra) so the set isn't accidentally lopsided.

Cross-tradition — Themes × Traditions

N=3 principles from all 12 traditions are arrayed across 26 themes. Each (theme × tradition) cell records whether the tradition attests to that theme, with N/D-attestation breadth scoring. Convergence is preserved; divergence is mapped, not smoothed.

Claim vs. warrant

The core analytic discipline: distinguish what a tradition claims from why it claims it.

Most traditions assert that every human being has worth. That's a shared claim. But the warrants diverge:

• Imago Dei — humans bear the image of God (Christianity, Judaism)
• Ātman = Brahman — the self is identical with the ultimate (Hinduism / Vedanta)
• Shared suffering without self — all sentient beings desire happiness; there is no enduring self to elevate (Buddhism)
• Ren — humanness is cultivated through right relationship (Confucianism)
• Alignment with the Dao — worth emerges through harmony with the way (Taoism)
• Khalīfah — humans are vicegerents on earth, accountable to God (Islam)

A shared claim with incompatible warrants is convergence of attention, not of belief. The Atlas preserves the distinction at every theme.

N/D-attestation breadth

Themes are scored by how many traditions attest (N) out of the total considered (D), then assigned a tier:

• UNIVERSAL — attested across all 12 traditions
• MAJORITY — 9-11 traditions
• MODERATE — 5-8 traditions
• WEAK — 2-4 traditions (still significant — pluri-tradition convergence, not noise)

Tier badges appear on every convergence theme. The same scoring is applied transparently — no manual selection of "important" themes.

Source traceability

Every principle traces back to specific verses or passages in the tradition's source text. This is maintained at every layer:

• N=1 statements cite specific verses (Dhp 5, Quran 2:62, BG 2.47, Jn 1:1)
• N=3 principles aggregate the N=1 evidence and re-cite
• Cross-tradition themes link back to N=3 principles per tradition
• Audit files (quote-audit-*.md) verify quote accuracy against source editions

Where the original-language scripture exists in public-domain translation, that translation is named (Müller for Dhammapada, Pickthall for Quran, JPS 1917 for Tanakh, etc.).

Reproducibility

The full methodology guides, per-tradition reviewer packets (r1-reviewer-packet.md), quote audits (quote-audit-*.md), and deep audits (audit-deep-*.md) are all in the public repository. The prompts used for AI-assisted extraction are documented. The architecture decisions (ADRs) are recorded.

The Atlas is one structured reading, not a final word. Where a within-tradition reviewer was unavailable, this is stated explicitly. Where evidence is partial, this is flagged. Where a held tension cannot be smoothed, it is left held.