Crown Jewel
When a Skeptic Designs the Test
We debated an AI skeptic (Grok, built by xAI) across multiple rounds about whether the Bible's coherence requires supernatural explanation. Grok dismissed all findings as "tradition + redaction + AI bias." So we challenged it to commit to specific falsification tests — and then ran them.
For non-technical readers: We asked a skeptical AI to design its own test — choosing 20 literary dimensions with zero religious content — and it committed publicly that if the Bible scored above 90% cross-author coherence, it would acknowledge the coherence is real and intrinsic to the text itself. The Bible scored 99.7%. The skeptic's own test confirmed what 25 other methods already showed.
Pre-500 BCE Embeddings
The challenge: Train Word2Vec on ONLY pre-exilic Hebrew text (Torah + early prophets, ~10,000 verses written before 500 BCE). Does the theological signal exist before any later tradition, redaction, or canonization process could reinforce it?
The result: Yes. The core theological-linguistic structure is fully formed in the oldest stratum alone. Covenant↔chesed similarity: 0.555. The lamb→sacrifice cluster is fully present. Atonement→sin/offering connections are detectable and coherent.
Grok's public acknowledgment
"The core theological-linguistic signal is already present and detectable in the pre-500 BCE corpus alone. That is a significant, noteworthy finding."
Skeptic-Defined Dimensions
The challenge: We told Grok to define its OWN 20 literary dimensions — with zero theology, zero Christianity, zero religious content of any kind. Pure literary analysis only. Then we'd score the Bible on Grok's dimensions and measure cross-author coherence.
Grok's 20 Non-Theological Dimensions
Zero theology. Zero Christianity. The skeptic's own dimensions.
Grok's public commitment
"If the Bible exceeds 90% cross-author coherence on these dimensions, I will acknowledge publicly that the coherence signal is demonstrably intrinsic to the text itself."
Coherence
Coherence
Coherence
Coherence
The inversion still holds. Cross-author coherence (98.9%) exceeds within-author coherence (98.0%). Authors are more consistent with each other than with themselves — even on purely literary, non-theological dimensions designed by a skeptic to avoid any religious signal. Jeremiah and Luke, separated by 600+ years, score 100%. Paul and Peter, who publicly disagreed, score 99.9%.
Why This Matters
Every prior test on this site used our dimensions. A skeptic could always say: "You designed the test to find what you wanted." This test was designed by a skeptic, using purely secular literary criteria, with a pre-committed threshold of 90%. The Bible didn't just pass — it scored 99.7%. The coherence isn't in our methodology. It's in the text.
Robustness Check
Genre-Controlled Analysis — Does the Inversion Survive?
A skeptic might argue the within < cross-author inversion is just because different authors write in different genres. Moses writes law AND poetry AND narrative — of course his within-author similarity is low. But when you compare only narrative authors to other narrative authors, does cross-author still exceed within-author?
The Answer
Yes. The inversion survives genre control at +1.5 percentage points. Even when we compare only storytellers to storytellers, only poets to poets, only letter-writers to letter-writers — Bible authors still sound more like each other than like themselves.
By Genre
| Genre | Within | Cross | Gap | Pattern |
|---|---|---|---|---|
| Narrative (7 authors) | 77.2% | 81.0% | +3.9pp | INVERSION |
| Gospel (4 authors) | 89.1% | 89.4% | +0.3pp | INVERSION |
| Poetry (5 authors) | 83.6% | 79.2% | -4.3pp | Normal |
| Prophecy (4 authors) | 84.8% | 83.1% | -1.7pp | Normal |
| Epistle (5 authors) | 84.4% | 82.9% | -1.5pp | Normal |
Cross-Text Comparison
Same test, three sacred/philosophical traditions. Genre-controlled within vs. cross-author similarity.
| Text | Within | Cross | Gap | Pattern |
|---|---|---|---|---|
| Bible (genre-controlled) | 80.7% | 82.2% | +1.5pp | INVERSION |
| Vedas + Upanishads | 88.8% | 58.9% | -29.9pp | NORMAL |
| Greek Philosophers | 99.1% | 96.9% | -2.2pp | NORMAL |
The 30-Point Gap
Crossing from Rigvedic hymns to Upanishadic philosophy
OT and NT storytellers sound more alike than each sounds like himself
For non-technical readers: We compared apples to apples. Instead of comparing all Bible authors to each other (which mixes different writing styles), we compared only storytellers to other storytellers, only poets to other poets, only letter-writers to other letter-writers. Even with this strict control, the Bible's storytellers sound MORE like each other than like their own other stories. And when we ran the same test on Hindu scriptures and Greek philosophers — both showed the normal pattern. Only the Bible inverts.
The Vedas gap is especially telling. Crossing from Rigvedic hymns to Upanishadic philosophy causes a 30-point DROP in similarity. That is what happens when different authors in different centuries write about different things. The Bible, crossing from OT narrative to NT narrative — different authors, different centuries, different languages — shows an INCREASE. The genre-mixing objection doesn't just fail. It makes the Bible's anomaly more visible.
Discovery 3
Dimension Pair Analysis
Which theological themes travel together? When one dimension rises in a passage, which others rise with it?
For non-technical readers: A correlation of 1.0 means two themes always appear together. A correlation of 0 means they're unrelated. When two dimensions have high correlation, it means the Bible consistently treats them as inseparable — even across different authors and centuries.
The Bible treats substitutionary sacrifice and blood atonement as inseparable (r=0.80). But firstborn reversal and rejected stone also travel together (r=0.68) — a connection theologians rarely make explicit. The youngest son who shouldn't inherit and the stone the builders rejected are the same story told two ways.
Discovery 4
Theological Outliers (Genre-Breakers)
Passages that break their genre's rules. A law passage scoring 10 on grace. A war story scoring 10 on suffering servant. These are the passages that don't fit — unless the whole book has one author.
For non-technical readers: "Deviation" measures how far above the genre average a passage scores. A narrative passage averaging 1.8 on blood atonement but scoring 10 has a deviation of 8.2 — it's doing something its genre almost never does.
Discovery 5
Genesis-Revelation Mirror Map
The Bible's beginning and ending are mirror images — computationally verified. These are the top 20 Genesis↔Revelation passage pairs by thematic similarity.
For non-technical readers: We compared every Genesis passage's 20-dimension fingerprint against every Revelation passage's fingerprint. A similarity of 0.95 means two passages — written roughly 1,500 years apart — are thematically near-identical.
Discovery 6
Theological Transition Points
Where does the Bible's theological center of gravity shift? These are the biggest dimension shifts — the exact passages where a theme suddenly appears or disappears.
For non-technical readers: "Shift" is the difference between the average score before and after a passage. A shift of +4.55 means a dimension was nearly absent, then suddenly became dominant. These are the narrative hinge points of Scripture.
Discovery 7
Author Blind Spots & Complementarity
Each author's weakest dimensions — and who fills the gap. If the Bible were a human collaboration, you'd expect overlapping strengths and shared blind spots. Instead, every weakness is precisely compensated.
Moses rarely writes about the Rejected Stone. Luke fills that gap perfectly. Moses is weak on Serpent Conquered. Mark picks it up. No single author covers everything — but together, every dimension is covered.
Discovery 8
Lost Typologies
The algorithm's auto-discovered OT→NT connections that may not appear in standard commentaries. These are the top 30 passage pairs by thematic similarity.
For non-technical readers: A "typology" is when an Old Testament event foreshadows a New Testament event. Some are well-known (Passover lamb → Christ's crucifixion). The algorithm found many that aren't in any commentary — Samson's death maps to Peter's confession. Joseph's prison dreams map to Revelation's witnesses.
Discovery 9
Inter-Testamental Gap
Does the signal survive the 400-year silence? Between Malachi and Matthew, no Scripture was written. If the coherence is human, it should break here.
Discovery 10
The Calvinist/Arminian Question — Computationally Resolved
For 500 years, the Church has debated whether salvation is driven by divine sovereignty (Calvinism) or human free will (Arminianism). We scored every passage on both frameworks simultaneously and let the data speak.
For non-technical readers: For 500 years, Christians have debated whether God chooses us (Calvinism) or we choose God (Arminianism). We scored every passage in the Bible on both frameworks simultaneously. The result: both are present, both are real, and they don't oppose each other mathematically. The Bible holds them in tension — 129 passages score high on BOTH divine sovereignty AND human free will. The data suggests the debate exists because both sides are right about what they see, and wrong about what they deny.
Method
We defined 5 Calvinist dimensions (divine sovereignty, unconditional election, effectual calling, perseverance, human inability) and 5 Arminian dimensions (human free will, conditional election, resistible grace, apostasy possible, prevenient grace). All 843 passages were scored on all 10, blind — the model was told to let the text speak for itself. No theological thumb on the scale.
Computational Verdict
The Bible is 12% more Calvinist overall (8.56 vs 7.66 out of 50), but the two frameworks are statistically independent (r = 0.19). They don't oppose each other — they operate on separate axes. The centuries-old debate is a false dichotomy: the data shows both divine sovereignty and human free will woven through the same text, often in the same passage.
out of 50
out of 50
ratio
between camps
Passage Lean Distribution
129 "tension passages" score 10+ on BOTH Calvinist and Arminian dimensions simultaneously. The Bible doesn't pick a side — it holds both truths in tension. Divine sovereignty is present in 93% of all passages. Human free will is present in 87%. They coexist almost everywhere.
Old Testament vs New Testament
The New Testament is more of both — not more Calvinist at the expense of Arminian, but intensified on every axis.
Author Leanings
Ratio = Calvinist / Arminian mean. Above 1.15 = Calvinist leaning. Below 0.85 = Arminian leaning.
| Author | Leaning | Ratio |
|---|---|---|
| David | CALVINIST | 1.78 |
| John | CALVINIST | 1.59 |
| Paul | CALVINIST | 1.47 |
| Isaiah | CALVINIST | 1.36 |
| Moses | CALVINIST | 1.23 |
| Ezekiel | BALANCED | 1.11 |
| Peter | BALANCED | 1.08 |
| Hebrews | BALANCED | 0.96 |
| Luke | BALANCED | 0.88 |
| James | ARMINIAN | 0.51 |
| Solomon | ARMINIAN | 0.33 |
Most Calvinist Passage
Nothing can separate us from the love of God. Predestined, called, justified, glorified — the golden chain of salvation with almost no human agency language.
Most Arminian Passage
"The soul who sins shall die." Individual moral responsibility, the ability to turn from wickedness, and conditional promises — pure human agency theology.
Top Tension Passages
Where divine sovereignty and human free will collide in the same text.
Dimension Presence Across All Passages
Discovery 11
The Gospel in Film — Hollywood's Unconscious Theology
We scored 144 popular films on the same 20 Gospel dimensions used for the Bible. The result: Hollywood's most beloved stories unconsciously reproduce the Gospel's DNA.
For non-technical readers: The stories that move us most deeply — stories of sacrifice, redemption, the outcast becoming the hero — are the Gospel's story retold in different settings. Hollywood doesn't know it's preaching.
Suffering Servant
Blood Atonement
highest genre avg
Most common dimension in film: Suffering Servant (5.72 avg) — Hollywood loves the hero who suffers nobly. Most absent: Blood Atonement (2.90 avg) — secular culture kept the attractive Gospel and dropped the offensive parts.
Gospel Score by Genre
Tolkien was a devout Catholic who said The Lord of the Rings was “a fundamentally religious and Catholic work.” The algorithm scored it 149/200 on Gospel dimensions — the highest of any non-explicitly-biblical film. The data confirms what Tolkien always claimed.
Discovery 12
Five Theories of the Atonement — Which Does the Bible Actually Teach?
Christians have debated for centuries how Jesus' death “works” to save people. Five major theories compete: Penal Substitution (Jesus bore God's wrath in our place), Christus Victor (Jesus defeated the powers of evil), Moral Influence (Jesus' example transforms us), Ransom Theory (Jesus paid a price to free captives), and Recapitulation (Jesus re-lived and redeemed human history). We scored every passage on all five.
For non-technical readers: Christians have debated for centuries how exactly Jesus' death “works” to save people. We scored every Bible passage on five competing explanations. The answer: all five are real, all five are present, and the Bible uses different ones in different contexts. The cross is bigger than any single theory.
Computational Verdict
The Bible doesn't teach one atonement theory. It teaches a multi-faceted atonement that each theory captures a piece of. The debate between theories, like Calvinist/Arminian, exists because each side is measuring a real signal in a text that's bigger than any single framework.
Penal Substitution ranks #1 by average score (mean 1.67), but Christus Victor is present in MORE passages (53.3% vs 51.8%). The Bible uses all five — no single theory has a monopoly.
Tension Passages — Where Theories Collide
Passages scoring high on 4+ atonement theories simultaneously.
Coming Soon
Future Research
Planned analyses that push the computational theology frontier further.
Word2Vec Analogy Theology
What happens when you run king - man + woman = ? on Biblical Hebrew embeddings? Can vector arithmetic discover theological relationships the same way it discovers linguistic ones?
Anti-Dimension Scoring
What does the Bible deliberately avoid scoring on? The absence of a theme can be as meaningful as its presence. Mapping the "negative space" of Scripture's theological fingerprint.
Cross-Reference Dark Matter
Of the 340,000 known cross-references, which connections are missing? If two passages have near-identical dimension fingerprints but no known cross-reference, that's a candidate for undiscovered intertextuality.
Emotional Arc Time Series
Plotting each dimension as a time series across all 843 passages. Does the Bible follow recognizable narrative arc shapes (Vonnegut curves)? Do all 20 dimensions converge at the crucifixion?