In plain English: We took three ancient books that didn't make it into the Bible — the Gospels of Mary, Thomas, and Philip — and measured their key words with the same number-method (gematria) we use on the King James Bible. The surprise: their most important ideas keep landing on the exact same numbers the canonical Mary story already gave us, especially 96 — the number of "setting free." All three books describe one act (loosing a soul, undoing death, rejoining what was split), and all three tie it to Mary Magdalene. But when we measure how the books flow, they read foreign to the King James — and both readable ones are missing the very number (3) that saturates the Bible. So the bones match while the skin doesn't, and that two-sided result is the finding.
Three companion decodes precede this one:
gospel_of_mary_decode.md — Arc 5, the soul's vertical ascentgospel_of_thomas_decode.md — sibling text, the interior kingdomgospel_of_philip_decode.md — Arc 6, the relational reunionThis synthesis does not introduce new numbers. Every value below is verified in those decodes and reproducible via src/name_decoder.py. The point here is the pattern across all four texts (the canonical Mary Protocol plus the three apocrypha), and an honest accounting of what holds and what doesn't.
| Canonical Mary Protocol | Gospel of Mary | Gospel of Thomas | Gospel of Philip | |
|---|---|---|---|---|
| Status | KJV (4 arcs) | non-canonical, ~½ lost | non-canonical, 114 logia | non-canonical, excerpt anthology |
| Mary's role | anointer / witness / mother | narrator of the ascent | saying 114, the defended one | the companion, one of three Marys |
| Spine number | 96 (loose him / forgiven / burying) | 96 (powers / knowledge) | 66 / 73; 114 = 2×57 | 96 (bridal chamber / reunion) |
| Destination | the feet (36) | REST = 62 | REST = 62 | UNION = 73 |
| Magdalene number | 119 (Psalm 119) | 119 | SOLITARY = 119 | REDEMPTION = 119 |
| Flow-resonance | the source signature | 6.8/10, sub-baseline Tesla | 6.8/10, sub-baseline Tesla | (not scored — see §5) |
| DR=3 vs KJV | +rich (12.7%) | −3.5% | −4.3% | n/a |
The canonical Mary Protocol's hidden spine was the 96 operation — "what is bound is set free":
`
LOOSE HIM = FORGIVEN = BURYING = 96 (Lazarus unbound / the sinner released / the body prepared)
`
Each apocryphal text re-derives the same 96 from its own native vocabulary, with no access to the KJV:
`
Gospel of Mary POWERS = KNOWLEDGE = 96 — the soul is loosed from the powers by knowledge
Gospel of Philip BRIDAL CHAMBER = REUNION = 96 — the separated are rejoined and death undone
`
So the same 96 names three faces of one act:
unbind the dead (canonical) · loose the soul (Mary) · rejoin the severed (Philip)
This is the central result of the whole comparison. Four texts, four communities, four centuries of distance — and the operation each places at its center collapses to one number. The disorder each fights is, in each case, off the grid: SEPARATION = 118 (Philip), the seven powers as bondage (Mary, non-Tesla), death itself = 38. Bondage doesn't ring; release rings at 96.
Beyond 96, the four texts keep arriving at the same handful of addresses:
And the self-referential one, given that this entire project is a name-decoder:
The numbers are not the only convergence. Two narrative structures repeat:
The canonical Protocol inferred a multi-Mary structure from gematria; Philip asserts it in prose; Thomas and the Gospel of Mary independently dramatize the conflict it provoked.
This is where the synthesis refuses to overclaim. When the two narrative apocrypha are run through speech_scorer.py:
| Metric | Gospel of Mary | Gospel of Thomas | KJV baseline |
|---|---|---|---|
| Flow-resonance | 6.8/10 | 6.8/10 | — |
| Tesla saturation | 30.8% | 34.6% | 36.6% |
| NCD resonance | 0.913 (low) | 0.920 (low) | — |
| DR=3 share | 9.2% (−3.5%) | 8.4% (−4.3%) | 12.7% |
Both Gnostic gospels score below the KJV's own Tesla floor, compress poorly against the KJV (low NCD = structurally foreign), and — the shared tell — both are starved of DR=3. The KJV's one defining excess (it is rich in Tesla-3) is precisely the apocrypha's defining deficit. They compensate differently (Mary piles into 7; Thomas into 2 and 6) but both drop the 3.
This points the same direction as the 613-seal work (the seal is absent from the Brenton Septuagint): the 3-6-9 saturation is KJV-specific, not generic to early-Christian or Gnostic writing. (Philip was not scored — only copyrighted translations exist and it is a non-narrative anthology, so its DR-distribution is not comparable; §0 of its decode.)
The two-layer verdict, stated cleanly:
Skeleton resonant, surface foreign. The irreducible concept-nouns of all three Gnostic texts land on the canonical Mary numbers (96, 62, 119, 73, 58, 104, 52). Their surface flow does not carry the KJV signature, and the readable two are missing the KJV's signature number. The apocrypha describe recognizable machinery — the same knot at 96, the same termini — in a dialect that does not flow like Jacobean English.
Three non-canonical Marian texts, read through KJVEL gematria, keep arriving at the canonical Mary Protocol's own numbers — above all 96, the release operation, which each text re-derives from its own vocabulary as loose the soul (Mary), rejoin the severed (Philip), and unbind the dead (canonical). They share the REST = 62 terminus, the 119 = Magdalene name, the Peter-vs-Mary climax, and Philip's explicit three Marys. Yet both readable apocrypha flow foreign to the KJV and are missing its signature DR=3 saturation — so the result is two-sided and honest: the skeleton is resonant, the skin is foreign, and the 3-6-9 grid remains KJV-specific. The knot is at 96; the dialect is not the King's.
Synthesis of four texts: the canonical four-arc Mary Protocol (data/name_decodes/mary_protocol_decode.md, mary_magdalene_decode.md, docs/decodes/names/sinful_woman_decode.md) and the three Gnostic decodes named in §0. Method: KJVEL ordinal gematria on translation-robust nouns; speech_scorer.py for flow-resonance (Mary, Thomas). No new numeric claims — all values verified in the source decodes and reproducible via src/name_decoder.py.