# Could a modern author construct the TJ?

## Independent AI reverse-construction stress test

September 21, 2026 · A supplement to the [neutral PHoax audit](/tjr/phoax-independent-analysis/index.htm). Prepared at the site owner's request. This is an adversarial feasibility exercise, not a manuscript authentication, an independent human panel, or a claim of revelation.

**Bottom line:** The TJ is a sustained alternative narrative with meaningful structural work and some useful teachings. A competent modern author could plausibly construct such a work over many revisions; a modern AI-assisted team could plausibly do so more quickly. This bounded experiment produced several relevant repairs without retrieving Deardorff's criticisms, but it did **not** produce or successfully test a complete 36-chapter book. It therefore refutes neither all fabrication hypotheses nor all arguments for an earlier source.

On the evidence examined, modern composition or substantial modern adaptation remains the better-supported working explanation, with **moderate, qualitative confidence**, not a calculated probability. That judgment does not identify an author, establish deliberate deception, exclude every ancient contribution, or resolve the historical truth of every shared episode. The requested conclusion, “the TJ is not a fabrication,” is not warranted by this test.

The important revision to the earlier audit is narrower: constructive examples now support its warning that a later author can anticipate some textual criticisms. At the same time, reading the entire TJ makes it inappropriate to dismiss its composition as effortless substitution. Both points can be true.

## 1. What was actually tested

The strongest version of the challenge is demanding: preserve broad Matthew correspondences across 36 chapters, anticipate hundreds of serious criticisms without seeing the later critique, maintain a distinct philosophy and narrative memory, add worthwhile lessons, and explain why a later Matthew editor would alter the result. Three different standards must not be conflated:

- A recognizable alternative gospel with recurring ideas is a literary production task.
- A work that survives independently selected historical, linguistic, ethical, and continuity tests is a much harder evaluation task.
- An ancient source demonstrably used by Matthew requires dated evidence; literary excellence alone cannot deliver it.

The exercise addresses the first and a limited portion of the second. It cannot accomplish the third.

### The blind phase and its deviations

An initial local extraction attempt was aborted when an overly broad extractor displayed fragments of Deardorff commentary, including editorial-fatigue discussions. That agent was no longer eligible to make blind predictions. Its attempt was not sealed or counted as evidence. A fresh, context-free builder was then assigned only Matthew, with no project reading, criticism retrieval, or communication of the comparison findings.

The replacement builder read all 28 chapters and 1,071 verse markers of the public-domain KJV Matthew from [Project Gutenberg ebook 10](https://www.gutenberg.org/ebooks/10), using its [plain-text Bible](https://www.gutenberg.org/cache/epub/10/pg10.txt). This is a disclosed departure from the initially requested extraction of Matthew-labelled quotations in the comparison pages. Using the complete gospel avoids inheriting Deardorff's selection of which verses deserved discussion; it also changes the exact benchmark and translation.

The resulting [Phase 1 artifact](/tjr/additional/phoax-audit-2026-09-21/reverse-construction-phase1.md) was frozen before Phase 2 at **2026-09-21 06:56:14 UTC**. It contains a design specification, a chapter map, 30 prospective changes, 16 attempted passages or structured plans, anticipated failures, and effort estimates. Its body contains approximately 12,321 words. It was not changed during comparison.

```text
Frozen body SHA-256:
78fd8298a13cf99ef7560b31e32b4bf3a3d4a67c84a16489ffb2a195901f86e0

Whole-file SHA-256:
970484fdd7f56ed688d0356065c52fe930a88c7cd0f6b3cab82c6ee4a76b20f8
```

The body hash uses the convention stated in that file: UTF-8, CRLF normalized to LF, everything after the body marker, exactly one final newline. The seal provides local integrity and ordering, not an independently notarized timestamp.

Two further limitations matter. First, the builder designed a roughly 50,000-word **28-chapter** work, following Matthew's chapter numbering, rather than the requested 36-chapter work. It did not write that full work. We retain the mismatch as a benchmark failure, not quietly repair it after seeing the TJ. Second, a pretrained model cannot demonstrate that its training excluded TJ material, gospel criticism, or familiar harmonizations. “Blind” here means no retrieval or task-context exposure during construction, not an uncontaminated historical mind. The commission itself also encouraged repair-seeking.

### The comparison phase

After the freeze, the reviewer read the [hosted TJ contents](/tj/tj.htm), its introductory material, every numbered chapter, and its epilogue. The site's `part` numbers are **not** chapter numbers: `part1.htm` through `part36.htm` reach only chapter 9. The full reading therefore continued through `part121.htm`; numbered chapters extend through `part118.htm`, followed by epilogue material. Duplicate website fragments were not counted as independent textual repetitions.

The reviewer also examined the [comparison introduction](/tjr/mttjindx.htm), the [audit README](/tjr/additional/phoax-audit-2026-09-21/README.md), report, summary and ledger, the public neutral-audit pages, the full Matthew 14 and 21 comparisons, and directed comparison sections on sexual ethics, geography, the passion, and edition changes. This was not a fresh adjudication of every one of the 343 counted ledger units, nor an independent collation of every book Deardorff cites. In particular, statements attributed to Beare below remain **Deardorff's documented attributions** unless otherwise stated.

The comparison uses explicit criteria: continuity, long-range consistency, chronology/geography, preservation versus divergence, philosophical integration, distinctive teaching, handling of externally attributed criticism, translation/edition effects, and adverse evidence. The 30 frozen predictions provide a complete small comparison register. Its match labels were assigned after reading the TJ by this reviewer, not by blinded independent raters. They are descriptive, not a preregistered numerical success score.

## 2. What the blind builder actually anticipated

The builder chose a limited retrospective narrator, retained a religious world, and emphasized accountability, practical compassion, uncertainty, and restitution. It explicitly declined technological explanations, a new biological father, a secret cosmology, and an escape-to-India narrative. Those decisions are substantial differences from the TJ, not failed attempts that may be relabelled as successes.

“Shared mechanism” below means that an important proposed repair resembles the TJ's repair, not that wording or an entire passage was recovered. “Different remedy” means the same problem family received materially different treatment. Omission is not counted as constructive explanation. References identify the hosted English reading; edition qualifications follow in section 4.

| Frozen prediction | Actual TJ comparison | Assessment |
| --- | --- | --- |
| P01: clarify mnemonic genealogy counting | TJ 1 adds an extended celestial ancestry and preserves a genealogy; it does not supply the proposed counting explanation. [Chapter 1](/tj/tj_files/part5.htm) | Not reproduced. |
| P02: distinguish legal/social fatherhood from conception testimony | Joseph is a guardian, but Gabriel is a biological celestial father. [TJ 1](/tj/tj_files/part6.htm) | Broad distinction shared; decisive mechanism different. |
| P03: clarify flight, danger, ruler's death, return | Flight remains, but Herod Antipas changes his mind rather than dying at the required transition. [TJ 2](/tj/tj_files/part8.htm) | Different remedy with new historical questions. |
| P04: frame the wilderness panorama as visionary experience | A transport and celestial instruction replace the temptation sequence. [TJ 3–4](/tj/tj_files/part10.htm) | Not the predicted repair. |
| P05: provide for disciples' households and funeral obligations | The harsh burial saying remains; no comparable household-provision program appears. [TJ 8:27–28](/tj/tj_files/part30.htm) | Important miss. |
| P06: distinguish visible good deeds from ostentatious motives | The TJ also criticizes display and public piety. Matthew already supplies much of this distinction. [TJ 6](/tj/tj_files/part20.htm) | Broad common teaching; weak diagnostic value. |
| P07: narrow denunciations and remove contempt for classes of people | Collective hostility toward Israel is extensive. [TJ 10](/tj/tj_files/part38.htm) | Opposite ethical direction. |
| P08: distinguish reasonable preparation from consuming anxiety | TJ 6:53–54 explicitly endorses preparing for tomorrow. [TJ 6](/tj/tj_files/part23.htm) | Shared practical repair. |
| P09: distinguish discernment from condemning persons | TJ 7:1 qualifies judgment as false judgment. [TJ 7](/tj/tj_files/part24.htm) | Shared clarification at a general level. |
| P10: acknowledge the swine keepers' economic loss | The destruction remains without the proposed acknowledgement. [TJ 8:29–37](/tj/tj_files/part31.htm) | Not reproduced. |
| P11: limit healing reports to witnesses and give secrecy a practical reason | Fear of capture and torture explicitly explains secrecy in TJ 13:25. Healings otherwise retain strong narrator certainty. [TJ 13](/tj/tj_files/part43.htm) | Specific secrecy mechanism shared; witness discipline not shared. |
| P12: distinguish a first local mission from a later broader one | The initial instruction instead avoids Israel and goes to Samaritans and the world. [TJ 10:5–6](/tj/tj_files/part38.htm) | Same tension, different remedy. |
| P13: remove short deadlines and distinguish conflict from commanded violence | The sword becomes a sword of knowledge, but the TJ has detailed long-range prophecies and much hostile rhetoric. [TJ 10](/tj/tj_files/part38.htm) | Partial conceptual overlap, not a full ethical or timing match. |
| P14: preserve disciples' memory while explaining fear or self-interest | The Peter episode connects fear and failed trust, but repeated rebukes of the disciples' understanding remain. [TJ 16](/tj/tj_files/part57.htm), [TJ 17](/tj/tj_files/part59.htm) | Partial; no demonstrated global memory repair. |
| P15: recast unforgivable speech as persistent refusal to change | The corresponding blasphemy discussion is not retained in the Sabbath chapter. [TJ 13](/tj/tj_files/part43.htm) | Omission, not the proposed reconstruction. |
| P16: standardize third-day language and remove “three days and three nights” | TJ 18:62–63 retains both formulations. [TJ 18](/tj/tj_files/part63.htm) | Proposed repair not made; a timing issue still needs interpretation. |
| P17: make parables invite thought rather than conceal truth | TJ 15:14–15 explicitly makes parables a stimulus to independent thought, while retaining distinctions in hearers' understanding. [TJ 15](/tj/tj_files/part47.htm) | Shared teaching mechanism. |
| P18: mark John's death as a flashback and retain both feedings with memory | TJ 16:4 marks the retrospective account; the second feeding is absent. [TJ 16](/tj/tj_files/part55.htm), [TJ 18](/tj/tj_files/part60.htm) | Clear flashback match; opposite treatment of the second feeding. |
| P19: retain the Canaanite woman and let her correct exclusion | The corresponding scene is absent between TJ 17 and 18. [TJ 17](/tj/tj_files/part59.htm), [TJ 18](/tj/tj_files/part60.htm) | Not reproduced; deletion cannot receive credit for the proposed lesson. |
| P20: make leadership accountable and integrate restitution/protection | Peter is denied the role of foundational authority; much of the proposed positive accountability program is absent. [TJ 18:23–24](/tj/tj_files/part61.htm) | Different remedy. |
| P21: make twelve thrones symbolic rather than guaranteed to twelve individuals | The thrones promise is omitted in favour of teaching about repeated lives. [TJ 20](/tj/tj_files/part69.htm) | Removes the issue by a different route. |
| P22: one rider on a colt, its mother alongside | The hosted TJ chooses the adult donkey. Its earlier edition reportedly retained the ambiguity. [TJ 22](/tj/tj_files/part72.htm), [edition discussion](/tjr/mt21.htm) | Shared recognition of ambiguity; different animal and a crucial revision caveat. |
| P23: an already withered tree, no curse, no literal mountain-moving promise | The enacted curse disappears, but TJ 21:11–15 retains both supernatural capabilities as promises. [TJ 21](/tj/tj_files/part70.htm) | Partial removal, not the proposed remedy. |
| P24: reconstruct the wedding feast without an implausibly timed war or clothing penalty | No corresponding wedding-feast narrative appears between TJ 22 and 23. [TJ 22](/tj/tj_files/part75.htm), [TJ 23](/tj/tj_files/part76.htm) | Omission, not constructive anticipation of the new version. |
| P25: remove inherited collective guilt and avoid absolving Pilate | The TJ extends collective blame; Pilate recognizes innocence yet yields. [TJ 24](/tj/tj_files/part84.htm), [TJ 29](/tj/tj_files/part102.htm) | Opposite or non-matching treatment. |
| P26: distinguish immediate danger, indefinite future, and present duty | TJ 25 moves much prophecy to a roughly two-millennia horizon with specific predictions. [TJ 25](/tj/tj_files/part85.htm) | Broad timing concern shared; specificity and mechanism differ. |
| P27: let Judas's offer change the plotters' original avoid-the-feast plan | The TJ uses another betrayer, Juda Ihariot, and lacks that plotting exchange. [TJ 27](/tj/tj_files/part92.htm) | Different construction, not the frozen prediction. |
| P28: account for garden testimony and continuity of fear/hearings | Judas remains awake as witness in the garden. The TJ also openly supplies supernatural foreknowledge elsewhere. [TJ 28](/tj/tj_files/part94.htm) | Witness problem partly shared; global narrator rule differs. |
| P29: retain remorse and omit risen saints | A different betrayer's remorse is integrated before Pilate; the risen saints are absent. [TJ 29](/tj/tj_files/part100.htm), [TJ 30](/tj/tj_files/part104.htm) | Broad overlap, not recovery of the distinctive Juda/Judas design. |
| P30: label the guard story as a report, allow journey time, retain doubt | The TJ asserts paralysis of guards, physical recovery and later travels as events. [TJ 31](/tj/tj_files/part105.htm), [TJ 31](/tj/tj_files/part106.htm) | No comparable testimony restriction; different ending. |

This register does not justify a “percentage predicted.” The units have different specificity, several contain both matches and misses, and the coding is post-comparison. Most importantly, neither generic concern about consistency nor omission of a difficult passage is equivalent to reconstructing a distinctive TJ solution.

There is nevertheless a real result. The builder wrote an explicit flashback, motivated secrecy, distinguished preparation from anxiety, and made parables teach independent thought without retrieving the comparison ledger. It also identified issues whose actual TJ remedies differed. That is more evidence of constructibility than merely asserting that a clever writer could repair Matthew. It is much less than demonstrating an independently coherent TJ-equivalent.

## 3. The whole-work comparison

| Criterion | Strongest feature of the actual TJ | Limitation or counterevidence | What the benchmark changes |
| --- | --- | --- | --- |
| Narrative continuity | The Juda/Judas distinction is planted, recalled during betrayal and trial, and tied to who preserves the teaching. | It introduces a near-identical-name conspiracy and privileged knowledge, which make continuity possible but require additional assumptions. | The blind builder did not recover this architecture; it deserves more credit than a local word replacement. |
| Long-range consistency | Celestial parentage, a human-like god below Creation, spiritual development, survival, travel, and later restoration repeatedly reinforce one another. | Repetition can preserve a doctrine while unresolved tensions remain in free will, destiny, love, and inherited blame. | A concept ledger and revision plan are plausible production tools, not proof they were used. |
| Chronology and geography | Explicit transitions sometimes improve orientation; John's death is clearly retrospective. | Bethlehem followed by the same-day seaside scene and the birth narrative's Herod identification need explanation; tomb timing is not self-evidently settled. | Obvious sequence repair is within the blind builder's demonstrated repertoire; accurate historical geography was not tested. |
| Preservation and divergence | The Matthew-like scaffold remains recognizable despite extensive expansion. | Removing an episode, reversing a teaching, or adding a new premise may avoid an objection at the cost of less correspondence. | There is no free requirement to preserve everything and simultaneously remove every difficulty. |
| Philosophical integration | Learning, responsibility and development recur in teaching and narrative. | “Laws of Creation” often assert authority rather than independently establish the law or its moral application. | Coherent doctrine is constructible; sound philosophy requires a separate argument. |
| Distinctive teachings | Mistakes can aid development; preparation, inquiry, non-coercive teaching and environmental care are useful themes. | The corpus also prescribes coercive penalties and collective blame and sometimes treats dissent as ignorance. | Practical value neither proves antiquity nor eliminates contrary teachings. |
| External scholarly problems | Some TJ variants address problems attributed to Beare and other scholars, not solely to Deardorff. | A scholar's later publication date does not prove the issue was undiscoverable earlier; some objections are weak or have other explanations. | A few prospective repairs reduce the force of “a modern author would never notice this.” |
| Translation and editions | Comparison pages often provide German as well as English; modern revisions are openly discussed. | The supposed Aramaic original is unavailable; English mediation and mixed editions prevent secure retroversion or automatic pre-1978 dating. | The benchmark tests English narrative design, not ancient Aramaic composition or Greek dependence. |
| Adverse evidence | The TJ is not uniformly evasive or devoid of explanatory detail. | It contains factual, ethical and literary vulnerabilities of its own, plus possible modern literary connections. | Evaluation must charge for new defects, not count only avoided Matthean ones. |

### A connected construction, not just a list of patches

The most impressive construction is not an isolated aphorism. [TJ 14](/tj/tj_files/part46.htm) introduces the faithful Judas and hostile Juda; [TJ 27](/tj/tj_files/part92.htm) explains the planned false attribution; [TJ 28](/tj/tj_files/part96.htm) dramatizes it; [TJ 29](/tj/tj_files/part100.htm) joins the death, money, burial and rumor; the final travels preserve the scribal role. That changes multiple dependent scenes. A modern author would need planning or several revision passes to make the alteration hold together.

Likewise, the administrator's son who survives attempted killing in the [vineyard parable, TJ 22:37–43](/tj/tj_files/part75.htm), anticipates the book's own survival account. The earlier teaching helps explain why opponents later fear an escape. The [tomb narrative](/tj/tj_files/part104.htm) supplies a second entrance, helpers, treatment and an injury intended to permit survival. These additions form a mechanism, not merely the assertion that he lived. They are also narrative stipulations, not independent medical or archaeological evidence that such a survival occurred.

Four broad rewriting policies could generate many apparently separate improvements: make the teacher human rather than divine; replace salvation and ecclesial authority with knowledge and development; motivate secret or private events through witnesses or extraordinary perception; extend the prophetic horizon and explain later corruption. Reusing a policy is skilled composition. It also means that its many consequences are correlated, not hundreds of independent coincidences.

### Problems recognized by other scholars

The [Matthew 14 comparison](/tjr/mt14.htm) attributes the retrospective-sequence issue to Kümmel's 1966 work as well as later commentators. The blind plan independently made that episode a clear flashback. Consequently, a later Beare citation is not evidence that a pre-1978 reader needed access to a future book. This example is a genuine reduction in the force of an impossibility argument, although it does not date the TJ's wording.

Deardorff attributes the mismatch between Herod wanting John killed and then grieving at his execution to Beare. TJ 16 changes Herod's response, but another primary scholarly explanation is available: Mark Goodacre's [“Fatigue in the Synoptics”](https://www.markgoodacre.org/Q/fatigue.htm) argues that Matthew's alteration of Mark leaves a residue of Mark's more protective Herod. Whether or not one accepts that account, TJ priority is not the only source-critical explanation of the seam.

The [Matthew 27:1–12 comparison](/tjr/mt27.htm#Mt27.1-12) attributes several placement and money/field objections to Beare, then adds further objections. The TJ's reordered scene and distinct betrayer give the sequence intelligible continuity. Yet a report of returning money followed by a description of throwing it down need not be two separate returns, and narrative compression need not mention every act of picking up coins. The package contains observations of unequal strength. It should not be converted into eleven equally compelling, independent surprises.

The [Matthew 26:14–16 discussion](/tjr/mt26.htm#Mt26.14-16) gives Bethany as roughly twenty miles from Jerusalem, an error already identified by the neutral audit. That inflated travel problem must not be credited to the TJ. Similarly, “then” can organize narration without implying that distant actors knew a speech had just ended. Some criticism demands a literalism of Matthew that its proposed solution relaxes for the TJ.

Even a completely accurate list of externally identified problems would establish only that the problems were not invented by Deardorff. It would not independently endorse his TJ solution, his date of that solution, or his likelihood estimate. The critical test is the earliest verifiable wording of each proposed repair against the earliest available formulation of the problem.

### Could a later Matthew editor have motives?

Yes, conditionally. An editor committed to a saving death and resurrection could replace survival and Indian travel; one committed to a divine son could suppress the hierarchy of celestial humans and Creation; an ecclesial writer could invert the TJ's rejection of Peter's authority. These are intelligible possible motives, and the direction-from-TJ model should not be dismissed as psychologically impossible.

But a modern religious reformer has equally intelligible reverse motives: replace doctrines the reformer rejects, vindicate a maligned disciple, supply bodily survival, and establish the authority of a rediscovered teaching. The two accounts are not equally evidenced merely because each can be narrated; the point is that motives fitted after both texts are visible do not discriminate strongly by themselves. A direction-of-dependence argument needs asymmetric evidence—dated witnesses, shared distinctive mistakes, demonstrable translation constraints, or a source relationship more specific than ideological preference.

## 4. Where the TJ itself is vulnerable

These are test cases, not an exhaustive adverse ledger. A fair audit distinguishes a plain factual failure, an unresolved tension, an ethical disagreement, and an unverified historical claim.

**Chronology and geography.** [TJ 14:1](/tj/tj_files/part46.htm) places Jmmanuel in Bethlehem; [TJ 15:1](/tj/tj_files/part47.htm) then places the seaside episode on the same day. The [comparison itself recognizes the difficulty](/tjr/mt13.htm#Mt13.1) and considers loose transitional wording or imperfect memory. Those are possible readings, but they show that TJ consistency also sometimes requires interpretive allowance. [TJ 2](/tj/tj_files/part8.htm) names Herod Antipas in the Jerusalem infancy episode. Josephus assigns Antipas Galilee and Perea, not Archelaus's Judean territory; a visit is not impossible, but the narrative's implied political setting needs explanation. See [Josephus, Antiquities 17, especially 17.11.4](https://penelope.uchicago.edu/josephus/ant-17.html). Substituting one Herod can change a chronology problem without establishing historical accuracy.

**Internal tensions.** [TJ 18:47–52](/tj/tj_files/part62.htm) treats mistakes as learning and rejects their transmission as a burden to later generations, whereas [TJ 24:44–53](/tj/tj_files/part84.htm) places a continuing blood burden on Israel. One could distinguish causal consequences from just punishment, but the text needs to sustain that distinction rather than relying on the reader to provide it. [TJ 5:41–45](/tj/tj_files/part19.htm) qualifies whom to love and despise, whereas [TJ 32:38–40](/tj/tj_files/part107.htm) commends unconditional love. These might concern different levels of practice, but that is a proposed reconciliation, not an absence of tension. [TJ 21:16–19](/tj/tj_files/part70.htm) makes a person's will and fate self-determined, while the passion speeches repeatedly invoke a detailed destined course. Foreknowledge and agency need not be logically incompatible; the work nevertheless owes an account of their relationship.

**Nature and sexual ethics.** The hosted [TJ 12:9–11](/tj/tj_files/part41.htm) invokes the supposed absence of male same-sex sexual conduct among animals and prescribes coercive bodily punishment. The biological universal is false: a primary comparative study documents male same-sex sexual behavior across many mammalian species. [Gómez, González-Megías and Verdú, Nature Communications (2023)](https://www.nature.com/articles/s41467-023-41290-x). Animal behavior would not by itself settle human ethics even if the premise were true; consent, welfare and rights require their own reasoning. This is not merely a reader disliking an unfamiliar cosmology.

There is an essential edition qualification. The [Matthew 5:28 discussion](/tjr/mt5.htm#Mt5.28) reports that, after a 2002 objection, the later wording was changed from castration/sterilization to exclusion from the community, explained as a mistranslation of “cut off.” The hosted TJ retains the earlier punitive reading. No surviving Aramaic was produced here to adjudicate that explanation. Therefore this audit neither attributes the same bodily penalty to every edition nor treats a later correction as independent evidence of pre-1978 foresight. Exclusionary punishment also deserves its own ethical scrutiny.

**A specific modern literary lead.** In [TJ 16:9–15](/tj/tj_files/part55.htm), Herodias's daughter loves John, kisses his severed head, and associates love with a bitter taste. Oscar Wilde's [Salomé, in a 1907 English edition](https://www.gutenberg.org/files/42704/42704-h/42704-h.htm), already combines the daughter's desire, the severed head, a kiss and the bitter taste of love. The TJ changes important details, including her culpability and where she kisses him. This is a distinctive candidate parallel, not proof of direct borrowing. A useful next test is comparison of the earliest German TJ with contemporary German Wilde/Strauss texts and older traditions. English translation could strengthen a verbal resemblance that was weaker in German. The lead should be investigated, not inflated into a solved attribution.

**Modern-compatible ideas are not automatically anachronisms.** Technological skies, population warnings, a later restorer, reincarnation and eastern travel fit a modern alternative-religious setting, but no one of those themes proves a date. Long alternative Jesus narratives and eastern wisdom were available well before 1978: the inspected 1911 edition of Levi's [Aquarian Gospel](https://www.gutenberg.org/files/44073/44073-h/44073-h.htm) contains extensive Indian and Tibetan episodes. Its existence establishes an available genre, not that the TJ's author read it, borrowed from it, or shared its entire theology.

**The documentary and translation layers.** The [TJ introduction](/tj/tj_files/part1.htm) and [purported translator's letter](/tj/tj_files/part121.htm) report a discovery and the loss of the originals. These are claims within the transmission story, not an independently verified chain of custody supplied by this exercise. The epilogue's agreement with another modern revelation does not independently date an ancient manuscript. Neither should a translator's English “spirit,” “consciousness,” or biblical idiom be uncritically treated as the original author's word choice.

[Deardorff's edition account](/tjr/genuine.htm#gen5) explicitly describes substantial changes in detail between 1978 and 1992, further changes in 2007, and his rejection of 2011 as a basis for the comparison. It also reports an explanation involving a typist's use of a German Bible and later corrections supplied through claimed extraterrestrial information. This documents the proponent's acknowledgement of editing and possible Bible mediation; it does not independently verify his explanation. The donkey ambiguity reportedly corrected in 1989–1990 is an especially concrete warning. A successful reading in a later edition is not automatically a successful prediction made before the criticism existed.

## 5. Human and modern-AI feasibility

### Before 1978

The task would be substantial, but “substantial” is not “implausible for any human.” A writer with Matthew, notebooks, printed reference works, a governing philosophy and time could track motifs and revise an entire book. Historical novelists do not need modern software to maintain a cast, itinerary or repeated doctrine. A collaborator or editorial history is possible, but none is established merely by finding the work complex.

The sealed builder estimated roughly **800–2,000 person-hours and 6–24 months** for a skilled solo writer's polished 50,000-word alternative narrative, after 4–8 major cycles. Those are planning judgments, not measurements of any 1970s person and not evidence that a named person possessed the necessary skills. The 36-chapter requirement adds eight chapter boundaries and, more importantly, additional conceptual and plot obligations if the expansion resembles the TJ's ending. Splitting 28 chapters into 36 is easy; designing and stress-testing the added cosmology, survival and travel is not. No defensible conversion factor from the sealed estimate is available.

Anticipating every independently selected serious objection, including obscure historical or linguistic ones, would be much harder than improving conspicuous narrative seams. It is not demonstrated for either the hypothetical writer or the actual TJ. The actual target contains misses, omissions, disputed “problems,” and later corrections. One must estimate the difficulty of producing the observed work, not an imagined flawless version of it.

### Three 2026 AI conditions

These conditions must be kept separate. A model that has the critique ledger is taking an open-book repair test; it cannot provide evidence of anticipating that ledger. “Matthew alone” restricts supplied material and retrieval, not everything embedded in pretraining. The table describes a capable model with human editorial review, a continuity register and repeated drafts, not a one-shot prompt.

Defects below mean **material internal defects per 100 episode-sized evaluation units**: inconsistent facts, impossible transitions within the chosen story, broken narrator rules, or a substantive alteration mislabeled as preservation. They are not percentages of false historical claims, probabilities of forgery, or measured model performance. All ranges are speculative; the first row comes from the frozen design, and the others are explicitly post-comparison extrapolations.

| Inputs | Likely capability and residual difficulty | Provisional internal-defect range: first / revised draft | Planning effort, not measured runtime |
| --- | --- | --- | --- |
| Matthew alone | Can identify obvious chronology, motives and repeated-theme issues and compose a new philosophy. Distinctive TJ-like choices and obscure scholarship are not reliably recoverable. Sustained adversarial quality: high difficulty. | 20–45 / 5–18 per 100 units | Frozen estimate: 6–30 total model-hours plus 60–180 human-hours, 5–10 passes, 2–6 elapsed weeks for the proposed 28-chapter work. |
| Matthew plus searchable modern scholarship, excluding the Deardorff/TJ corpus | Can build a much broader criticism register and check geography and historical assumptions. Research can also import disputed consensus as if it were fact, or contaminate a purported pre-1978 test. High, but reduced discovery difficulty. | Indicative 15–35 / 3–12; no empirical calibration | Budget approximately 6–12 substantive passes and 80–220 human-hours; research/verification may consume time saved in drafting. No validated 36-chapter runtime. |
| Matthew plus Deardorff's critique ledger | Can target listed objections and make proposed editor motives explicit. Easiest condition for apparent ledger coverage, but globally consistent integration and independent held-out tests remain difficult. | Indicative 10–30 / 2–10; overlaps other ranges substantially | Budget approximately 6–12 passes and 80–200 human-hours, with a separate held-out review. These numbers do not predict near-zero failure on every criticism. |

The difference between conditions is primarily access to the target, not a guarantee of good prose or fewer unforeseen defects. Longer instructions can themselves cause omissions, incompatible fixes and overfitting. A primary 2026 study, [ConStory-Bench](https://arxiv.org/abs/2603.05890), reports persistent factual and temporal consistency problems in long model-generated stories, including interacting failures. It supports caution about whole-book coherence; it does **not** validate the numerical ranges above or test this commission.

A serious 36-chapter production workflow would maintain a dated episode graph, travel and age chronology, who-knows-what register, vocabulary definitions, doctrine-to-scene links, and an explicit matrix of each retained, altered or omitted Matthew episode. Every repair would be followed through its consequences. Separate reviewers would inspect geography, narrative voice, ethics, historical plausibility and source dependence. These are post-comparison workflow recommendations, not additions to the sealed predictions.

The likely outcome is a plausible alternative narrative with a nontrivial residual defect burden, not an error-free ancient source. Zero defects across “hundreds of serious criticisms” is an unperformed test, not a capability claim justified by this report.

### Why present-day imitation does not solve the historical question

A 2026 model benefits from large-scale training, searchable texts, automated checking and rapid parallel consideration of alternatives. Its successful imitation could establish current feasibility under stated inputs; it could not show what a particular pre-1978 person knew, which books that person read, how long that person worked, or when the earliest TJ wording existed. Conversely, an AI's failure would not establish human impossibility. Historical writers may bring enduring plans, cultural experience and literary discipline that a single model run lacks.

The blind samples therefore weaken a universal impossibility claim, not a specific historical alibi. Historical attribution still requires historical evidence.

## 6. Authorship, money, fame and other motives

The proposition “no one claimed authorship, therefore the author obtained neither money nor fame” combines several different claims. They need separate evidence.

| Claim or consideration | What this review can establish | Consequence |
| --- | --- | --- |
| No identified modern claimant says “I fabricated it” | No such admission was found in the material reviewed. The book instead presents ancient authorship and modern discoverers/transmitters. | Absence of an admission is unsurprising under both sincere ancient-source belief and deliberate pseudepigraphic presentation. It is not strong discrimination by itself. |
| No money was earned | This review has no audited sales, royalties, costs, donations or personal financial records. | Neither profit nor absence of profit is established. A book's existence or availability is not a balance sheet. |
| No fame or reputation accrued | Meier is publicly named as discoverer/editor in the transmission story. TJ 14–15 predicts later restoration of the teaching. | The absence of a novelist's byline does not rule out reputation as discoverer, custodian or interpreter. This identifies a possible mechanism, not demonstrated intent. |
| A purely cash-seeking or literary-celebrity motive | Verified lack of material benefit or authorial credit would count against these narrow motive models. | It would not eliminate all modern-composition models. |
| Ideological, religious, community, experimental or collaborative motives | Such motives are possible alternatives; this review supplies no private correspondence or testimony proving them. | Keep them as hypotheses, not allegations about named people. |

A writer can value spreading a worldview, correcting a religion, influencing a community, or gaining credibility by proxy more than royalties or public literary credit. A person can also sincerely believe a text received through unusual means to be true. “Modern composition” and “deliberate hoax” are therefore not identical categories. Collaboration may distribute incentives, but invoking an invisible team without evidence is no more an explanation than invoking an impossibly gifted unknown solitary writer.

If anonymity and lack of benefit were independently established, they would lower the plausibility of some prior motive models. They would not supply a numerical prior for ancient authenticity, and they would not outweigh direct dating evidence. In this case the necessary financial evidence is missing, and the named transmitter's public role makes the strongest “no possible fame” premise too broad. No accusation of fraud follows from that observation.

## 7. Reassessing PHoax and the earlier audit

The [earlier neutral report](/tjr/phoax-independent-analysis/index.htm) correctly distinguished the **source of a Matthew criticism** from the **author of a TJ interpretation and score**. It counted 343 units. Applying Deardorff's existing inputs produced about 1.64 × 10^-110; the 193-unit external-scholar subset produced about 8.25 × 10^-61. These extreme outputs describe his scoring model, not independently observed frequencies of successful forgeries.

Its separate 36-unit AI review yielded 7 TJ-favouring, 27 neutral and 2 modern-derivation-favouring assessments. Dependence-aware grouping gave a diagnostic index of about 0.458, with a sensitivity range of 0.201–0.739. That was not a measured posterior, a confidence interval, or a representative human vote. The reverse-construction exercise does not turn it into one.

The central statistical problem remains. Multiplying passage-level odds only produces the claimed evidential accumulation if inputs have appropriate likelihood meaning and dependence is handled correctly. One authorial policy can explain dozens of related changes; the same selected critic and evaluator can shape many entries. A bounded constructive match is not an empirically calibrated replacement value for a ledger score. We therefore **leave the original ledger and numerical outputs unchanged** and publish this as an explicit interpretive reassessment.

### Strongest evidence on each side

The strongest case for taking a pre-Matthean or otherwise earlier TJ-like source seriously is the existence of connected expansions that explain more than one local seam, the coherence of a sustained alternative framework, and the fact that some relevant Matthew problems were identified outside Deardorff's project. These features make a simplistic dismissal intellectually inadequate. They justify careful source-critical study.

The strongest case for modern composition or adaptation is documentary and comparative: no independently examinable ancient witness was established here; significant modern revision is acknowledged; some apparent repairs are not demonstrably pre-1978; ordinary editing explains several blind matches; the TJ retains or introduces difficulties; and its distinctive modern-compatible narrative has plausible earlier modern literary and religious comparators. The Salomé parallel is a lead, not the foundation of the conclusion. The judgement would still require caution without it.

The lack of original material does not prove fabrication—many genuinely ancient works survive only in copies—but it prevents using palaeography, materials, language and custody to break the symmetry in this particular claim. The relevant asymmetry is not “unusual beliefs are false”; it is that the ancient-source claim needs evidence independent of the modern text and its own provenance narrative.

### Four conclusions that should remain separate

| Question | Revised assessment | Confidence and boundary |
| --- | --- | --- |
| Is the TJ textually sophisticated? | Yes: it shows sustained motifs and several connected narrative revisions, alongside uneven execution. | Moderate to high confidence as a literary judgment; not a measurement of unique genius. |
| Does textual comparison establish which text depends on which? | No. Some TJ expansions explain Matthew, but later expansion, common tradition and mixed editing remain live explanations. | High confidence that this benchmark does not decide direction; no claim that every model fits equally well. |
| Is the extant TJ demonstrably ancient? | Not established by the reviewed evidence. Modern editing is positively documented; modern composition or substantial adaptation is the leading working model here. | High confidence about the missing authentication and editing; moderate qualitative confidence about the preferred model. No identified fabricator or exclusion of every ancient core. |
| Are its extraordinary historical or supernatural claims true? | Literary continuity and useful teachings do not verify celestial visitors, reincarnation, survival or a revelatory restoration. | This exercise does not adjudicate those claims by experiment or independent historical proof. |

The exercise changes the argument about **difficulty**, not the date of a manuscript. It supports “a nontrivial modern construction is feasible” more than “the actual text was certainly fabricated.” It does not support “too good to be fabricated,” and it does not support “successful imitation proves fraud.”

## 8. Falsifiers and the next decisive tests

1. **Establish the earliest text.** Obtain complete, independently dated scans of the 1978 German edition and any verifiable earlier typescript; collate them with 1992, 2001 and 2007. Record each claimed repair, addition and retranslation separately. A later correction must not inherit an earlier publication date.
2. **Audit the provenance independently.** Locate original correspondence, independently identifiable witnesses, publication records and material evidence with a documented custody chain. An authenticated ancient TJ-specific witness would strongly change this assessment; a demonstrably modern shared error or securely dated source borrowing would strengthen the adaptation model.
3. **Build a problem register before looking up TJ solutions.** Have scholars select and characterize Matthew problems from primary research, record pre-1978 availability, distinguish contradiction from compression and interpretation, and retain failures and disputed cases. Verify Beare and other citations against the books, not only this site's summaries.
4. **Run the full construction test.** Commission complete 36-chapter works under all three input conditions, with several independent builders, fixed time budgets, preserved drafts and genuinely held-out criticism. Human participants should have documented source access; model participants need disclosure of pretraining limitations. Raters should not know which work or condition they are scoring.
5. **Score integration, not only avoidance.** Penalize new contradictions, arbitrary omissions, lost Matthew correspondence, unethical proposals, invented historical certainty and inconsistent narrator access. Report agreement between raters and all failed attempts. Do not treat a missing passage as an ingenious solution unless the rubric explicitly rewards omission.
6. **Test discriminating parallels.** Compare the John/Herodias scene in early German TJ, Wilde and Strauss; test distinctive shared errors and wording against translation alternatives. Examine whether alleged Semitic features exceed what biblical German or English naturally produces. A hypothetical reconstructed Aramaic is not a manuscript witness.

These tests can change the conclusion in either direction. Merely finding more agreeable teachings, more plausible editorial motives, or another modern statement repeating the discovery story would not resolve provenance.

## 9. Which text would I use for practical guidance?

**Both with critical controls, with Matthew carrying more weight for interpersonal ethics.** That is a reading recommendation, not a revelation claim or a judgment that canonical status establishes truth. If restricted to one for everyday moral practice, I would choose Matthew read critically.

Matthew's enemy-love, mercy, resistance to ostentation, and attention to hungry, sick and imprisoned people are demanding moral resources (Matthew 5:43–48; 6:1–4; 25:35–40). Those strengths are not reducible to naive passivity or merely “basic doctrine.” Its hostile collective rhetoric, punitive judgment language and self-mutilation imagery require equally explicit controls; neither inherited guilt nor literal bodily harm should be adopted as a practical rule.

The TJ is worth reading for preparation rather than anxious neglect ([6:53–54](/tj/tj_files/part23.htm)), learning from mistakes ([18:47–52](/tj/tj_files/part62.htm)), examining even the teacher's claims ([19:14–15](/tj/tj_files/part65.htm)), refusing forced belief ([32:13](/tj/tj_files/part107.htm)), and environmental responsibility ([chapter 36](/tj/tj_files/part116.htm)). Its distinction between an admired teacher and one's own responsibility can be valuable even without accepting its cosmology.

The controls are indispensable: reject coercive sexual penalties, inherited collective blame, the idea that nature automatically dictates a just social policy, and promises of extraordinary power used to blame suffering on insufficient knowledge or trust. Read the universal-love passages alongside the exclusions, not instead of them. Claims about healing are not a substitute for healthcare, and dissent is not evidence of moral inferiority.

The ethical test should be consequences, compassion, consent, fairness, intellectual honesty and willingness to revise—not which text wins an authenticity contest. A genuine ancient text can contain bad guidance; a modern text can contain excellent guidance. Neither fact dates it.

## 10. Reproducibility and scope of the result

The frozen Phase 1 artifact, its hashes, the complete 30-prediction comparison register, and the linked source pages preserve what was attempted and what was subsequently observed. No source TJ or Matthew-comparison text was changed. The existing scoring ledger was not rescored. The public digest uses the neutral audit's visual system and links to the complete report in the site's Markdown reader.

The result remains limited by one procedurally blind builder, pretrained knowledge, a 28-versus-36-chapter mismatch, no complete generated book, one unblinded comparison reviewer, mixed translation/edition layers, and no independently authenticated ancient witness or financial investigation. Labor and defect estimates are deliberately identified as conjectural planning ranges. The historical and philosophical judgments are reasoned assessments, not experimental proof.

**Final judgment:** The TJ's literary work merits serious examination, but its coherence and useful lessons do not make modern construction implausible or authenticate an ancient source. The reverse exercise modestly strengthens the case that many repairs can arise through ordinary, sustained authorship; it also demonstrates why the whole-book challenge deserves more than a casual dismissal. Modern composition or substantial adaptation remains the preferred provisional explanation on this record, without a finding of a particular person's fraud.