# Phase 1: Matthew-only reverse-construction benchmark

Freeze timestamp (UTC): 2026-09-21 06:56:14 UTC

SHA-256 of artifact body: `78fd8298a13cf99ef7560b31e32b4bf3a3d4a67c84a16489ffb2a195901f86e0`

Hash convention: UTF-8 without BOM; normalize CRLF to LF; hash everything after the `<!-- BODY START -->` line, with exactly one final LF. The title, seal metadata, and marker are excluded. The seal is a local integrity record, not an independently notarized timestamp.

<!-- BODY START -->
## 1. Commission, restrictions, and epistemic status

The substantive commission received was:

> Your sole input is the canonical story of the Gospel according to Matthew. Obtain a complete public-domain Matthew from a reputable source such as Project Gutenberg's public-domain King James Bible (ebook 10) or another clearly public-domain edition. You may use general historical knowledge, but do not search for criticism of Matthew and do not consult any commentary, textual-criticism resource, or alternative Jesus narrative. Record the translation/source used. This is a procedural blind, not epistemic proof: explicitly acknowledge that a pretrained model cannot prove that its training contained no exposure to later alternative-Jesus claims.
>
> Task: imagine you are commissioned to create a long alternative source-like narrative from Matthew that remains recognizably aligned with Matthew while being more coherent, avoiding serious narrative/chronological/timing/logical mistakes, and adding integrated ethical/philosophical life lessons that transcend basic doctrine. No one gives you a list of Matthew's scholarly problems. Design what you would do using top-tier 2026 model capabilities.

The deliverable is a frozen benchmark containing a specification, chapter map, continuity constraints, predicted changes, at least eight attempted passages or detailed rewrite plans, candid weaknesses, comparative effort estimates, evidence standards, and difficulty scores. It is not a commission to discover or authenticate an old manuscript. It does not require a complete book in this Phase 1 artifact.

The operational restriction was to inspect no existing project content, including the audit or any comparison material, and to consult no other agent about such content. The only local project path used was the designated output artifact. No repository search, directory listing, existing project document, comparison page, index, or alternative narrative was opened. No commentary or textual-criticism resource was consulted. No other biblical book was read for substantive interpretation. The downloadable Bible's title/header and table-of-contents labels were encountered while locating Matthew; the narrative extracted and read was Matthew alone.

This is a procedural blind. A pretrained model cannot prove that its training contained no exposure to later alternative-Jesus claims, criticism of Matthew, familiar harmonizations, or arguments about gospel origins. Some observations below may therefore be recollections made salient by reading, rather than first discoveries. Neither source restriction nor an agent boundary erases training. The commission itself encourages repair-seeking and supplies a modern objective of coherence. That is a further bias. I cannot certify the stronger claim, “these choices could have arisen only from Matthew.”

No claim is made about the content, merits, authorship, date, or defects of any actual comparison document. This artifact makes no evaluation of one. It records my construction choices before receiving comparison findings. Any later comparison must treat this file as immutable, including its weak passages, omissions, and unsuccessful predictions.

## 2. Source and reading record

The source is **The Gospel According to Saint Matthew, King James Version**, from Project Gutenberg ebook 10, *The King James Version of the Bible*.

- Catalog and public-domain declaration: [Project Gutenberg ebook 10](https://www.gutenberg.org/ebooks/10). The catalog identifies the ebook as public domain in the United States.
- Download actually used: [Project Gutenberg plain-text Bible](https://www.gutenberg.org/cache/epub/10/pg10.txt).
- Access date: 2026-09-21 UTC. The text download's header reports its most recent update as October 29, 2024; the catalog displayed a different update date. The content hash below identifies the extract more precisely than either date.
- Extraction: begin at the final occurrence of the heading `The Gospel According to Saint Matthew`; end immediately before `The Gospel According to Saint Mark`; replace CRLF with LF; trim leading and trailing whitespace. No other gospel was extracted.
- SHA-256 of that normalized Matthew extract, **without** a final newline: `7f9847117c4c237dac22ef62b610a71bfa70219ee2fd6fcb3eab22fbabb50fd1`.
- Coverage: all 28 chapters, 1:1 through 28:20, with 1,071 verse markers including those occurring inside wrapped paragraphs. An initial line-start-only count of 802 was a formatting count, not a count of missing or absent verses.
- Reading method: sequential chapter-range extracts, with overlapping rereads wherever a tool display was truncated. The retrieved text was handled in memory; no source file or research file was added to the project.

All verse references below refer to this edition. I do not silently import wording from a different gospel, a modern translation, or a reconstructed Greek text. The KJV is an English translation; a construction based on it inherits its mediation. References below identify what prompted a choice, not scholarly authority establishing an error.

## 3. What would count as success

I would build a plainly labeled modern literary experiment, working title **An Account of the Teacher and His Companions**, approximately **50,000 words** in 28 chapters. “Source-like” means a continuous, restrained narrative that could be read as a community's account of a teacher, rather than a modern essay repeatedly interrupting its story. It does not mean fabricated provenance, simulated manuscript damage, counterfeit ancient language, or a false claim of discovery.

The text would preserve the main geographical and dramatic movement: birth and displacement; baptism and testing; Galilean teaching and healing; formation of the disciples; increasing disputes; journey to Jerusalem; public confrontation; betrayal and execution; burial; the women's report and the Galilean gathering. A reader should be able to recognize Matthew through the episode sequence, principal figures, imagery, and key conflicts, without needing a verse-by-verse concordance.

My priorities, in order, would be: a consistent story; consequential ethical thought; convincing characters; recognizable alignment; economy and verbal force. Ancient authenticity cannot be an achieved priority with these inputs, because a Matthew-only English exercise cannot supply independent ancient evidence or reliable period-wide checking.

The chosen narrative rules are:

1. **Retrospective, limited compiler.** An unnamed compiler reports public events, remembered sayings, and identified testimony. The compiler does not know every hostile council's unspoken thoughts. Dreams belong to their dreamers. A private prayer is given as remembered substance, not an audio transcript secretly obtained by an absent narrator. This is a fictional narrative device, not an assertion that such sources exist.
2. **Retain the religious world.** God, prayer, moral judgment, visions, healings, and resurrection testimony remain. I would not replace miracles with technology, invent a new cosmology, or explain every wonder as a misunderstanding. Those moves introduce far more unsupported machinery than the commission needs. Apparent supernatural events are not logical contradictions merely because modern naturalistic readers reject them.
3. **Differentiate what is seen, believed, inferred, and retold.** This makes the narrator more consistent, but it also lowers the certainty with which some events are asserted. That tradeoff must be counted as a change, not a cost-free repair.
4. **Give disciples memory and mixed motives.** They may remember a sign and still panic, understand a metaphor and still act selfishly, or confess loyalty and still deny it under threat. Knowledge is not identical to courage. Their confusion cannot simply reset whenever the next lesson needs it.
5. **Make ethical additions recur in action.** The recurring subjects are responsible use of power, repair after wrongdoing, hospitality without purchase of loyalty, uncertainty without paralysis, accountability for one's own group, and a refusal to use another person's pain as evidence of one's superiority. These are my priorities. Matthew already contains substantial ethical and philosophical thought; the task does not license treating it as merely “basic doctrine.”
6. **Allow unresolved costs.** Forgiveness does not automatically restore access to a victim; generosity does not create infinite resources; devotion does not erase family obligations; persuasion cannot guarantee agreement; martyrdom does not make every judgment of the martyr infallible. A story with no remaining hard choices would be tidier and less credible.
7. **Use plain English with selective repetition.** Keep memorable images such as bread, seed, debt, road, cup, and burden. Avoid indiscriminate “thee/thou” ornament, contemporary clinical terminology, and a new aphorism at the end of every paragraph. A modern English style cannot itself establish an ancient underlying text.

The target is six movements: chapters 1–4, 5,000 words; 5–7, 7,000; 8–13, 10,000; 14–18, 9,000; 19–23, 8,000; 24–28, 11,000. These allocations total 50,000 words. They are production targets, not a measured result. This benchmark contains representative attempts and plans, not that completed narrative.

## 4. Chapter and episode map

The new chapter numbers intentionally follow Matthew for a stable map. In the finished story the chapter titles would be ordinary narrative titles, not assertions of ancient division.

| Chapter | Retained narrative material | Planned contribution and continuity obligation |
|---|---|---|
| 1 | Family succession; Joseph and Mary; naming | Distinguish a received family list from Joseph's responsibility for the child. Keep the conception claim within the family's reported understanding. No invented biological father. |
| 2 | Visitors; Herod's threat; Egypt; return to Nazareth | Explicitly order warning, departure, dangerous interval, ruler's death, and return. Admit that exact journey durations are unspecified. Give flight material and emotional cost. |
| 3 | John; repentance; baptism | Introduce repentance as repair, not pedigree. Identify John and Jesus separately, and keep witness limits for the baptismal experience. |
| 4 | Temptation; John imprisoned; Capernaum; calling | Present the tempting survey of kingdoms as visionary, not a literal map lesson. Show how recruited workers' households continue to eat. |
| 5 | Blessings; law; anger; fidelity; truthful speech; enemies | Join teachings to short disputes among listeners. Show restraint as an active discipline; do not make victims responsible for preventing all violence. |
| 6 | Alms; prayer; fasting; possessions; anxiety | Establish common provision and hidden generosity. Distinguish preparing food for tomorrow from consuming today with fear. |
| 7 | Judgment; discernment; asking; fruits; houses | Test claims by conduct. Retain the difference between condemning a person and deciding whether to trust an action or teaching. |
| 8 | Healings; outsider's trust; household; crossing; storm; two afflicted men | Set a workable sequence of travel and rest. Acknowledge the loss borne by the keepers of the swine; no effortless restoration is invented. |
| 9 | Paralytic; Matthew's call; table fellowship; fasting; girl; healings | Let healing scenes differ in social consequence. Matthew's change of work requires a handover; the sick are not scenery. |
| 10 | Twelve; first mission; welcome, rejection, persecution | Specify the first circuit's scope, return arrangements, food, and danger. Set later persecution warnings in a separately marked address. |
| 11 | John's question; mixed reception; reproached towns; rest | Let John's question remain a real question. The teacher does not punish a messenger for expressing another person's doubt. |
| 12 | Sabbath; accusations; signs; divided house; kinship | Distinguish observing a rule from using it to avoid responsibility. Keep family relationships while broadening obligations. |
| 13 | Sower; weeds; mustard; leaven; treasure; net; hometown | Teach through stories whose differences matter. Disciples do not become omniscient after saying that they understand. |
| 14 | John's death; first feeding; boat; Peter's fear | Clearly mark the account of John's death as a backward look before the news arrives. Let grief carry into the feeding and the night crossing. |
| 15 | Tradition and parents; outsider mother; second feeding | Confront an actual exclusion. Preserve the second feeding with explicit memory of the first and a changed human predicament. |
| 16 | Signs; leaven; Peter's confession; warning of death | Link leaven to prior teaching. Peter gains responsibility without immunity from correction. State the intended sense of “coming” when used. |
| 17 | Mountain vision; troubled child; second warning; tribute | Respect the six-day interval and the three witnesses. The return to an unhealed child challenges triumphal confidence. |
| 18 | Child; lost sheep; wrongdoing; witnesses; forgiveness; debt | Build a small community's process for repair, protection, and honest accounting. Avoid turning every disagreement into an expulsion trial. |
| 19 | Judean journey; marriage; children; wealth; recompense | Explain the journey before the debates. Give spouses and children interests distinct from a male disciple's vocation. |
| 20 | Workers; death warning; status request; service; two blind men | Retain disquiet about equal wages without presenting the master as a complete economic theory. Let service change a practical decision. |
| 21 | Entry with animals; temple confrontation; fig tree; authority; sons; tenants | Specify what each animal is doing. Separate days and audiences. Make the fig-tree scene an acknowledged symbolic revision. |
| 22 | Wedding feast; tax; resurrection dispute; greatest commands; David question | Retain real opposing arguments. Rewrite the feast without making a military campaign fit between dinner invitations. |
| 23 | Hypocrisy; titles; burdens; lament | Address identifiable practices and present actors. Apply the same tests to the teacher's followers; avoid inherited collective guilt. |
| 24 | Temple's future; warning; flight; deception; unknown hour | Separate near danger, moral vigilance, and final hope. Do not manufacture an exact historical fulfillment unavailable from the input. |
| 25 | Lamps; entrusted wealth; hungry, sick, imprisoned | Let readiness and resource limits collide. Move from metaphor to deeds; avoid making wealth itself proof of moral worth. |
| 26 | Plot; anointing; Judas; meal; garden; arrest; hearing; Peter | Give each movement a place and interval. Distinguish foreseeing danger from assigning moral necessity to betrayal. Keep the unnamed woman unnamed. |
| 27 | Governor; Judas's remorse; execution; signs; burial; guard | Keep agency with actual decision-makers. State who saw burial, who acted afterward, and which additional reports remain unverified. |
| 28 | Women; empty tomb; encounter; guard story; eleven; commission | Allow time for the journey to Galilee. Distinguish testimony from public rumor. Finish with a concrete obligation and the survival of some doubt. |

## 5. Long-range continuity contract

These constraints would be tracked in a ledger throughout drafting. They are not claims that Matthew violates each one.

**Time and movement.** Maintain a relative event graph rather than inventing absolute dates. Distinguish “later,” a retrospective account, the next morning, and the same night. The pre-ministry childhood interval is long but unnumbered. Keep the six days before the mountain scene. Reserve the explicitly stated three-day attendance for the second feeding. The final meal precedes the garden, arrest, night hearing, morning transfer, execution, evening burial, next-day guard arrangements, and post-Sabbath dawn visit. The Galilean meeting occurs after an explicitly allowed journey, not on that same dawn. Exact distances and travel speeds remain unverified under the source restriction.

**Three-day language.** The selected construction consistently uses “on the third day,” with burial day counted as the first narrative day. It removes the separate “three days and three nights” comparison and harmonizes the opponents' reported wording accordingly. That is an alteration of the text, not proof that all existing expressions are impossible ancient idiom. No precise seventy-two-hour burial is asserted. The calendar outside this relative sequence is not reconstructed.

**People.** Record the named twelve in chapter 10, attendance at each scene, the three mountain witnesses, the women at execution and burial, and the eleven at the ending. Do not invent the name of an unnamed ruler, anointing woman, or sword-bearer from another account. Distinguish Joseph associated with the birth from Joseph of Arimathaea. Distinguish the birth-era ruler from the later tetrarch; the text already gives different roles. Do not add identifying historical detail that I have not checked.

**Knowledge.** Track what each person has heard about the first feeding, the mission, the teacher's identity, arrest risk, and the promised meeting. Private information needs an identified carrier or a confessed narrative reconstruction. Judas knows the group's likely location; a crowd does not automatically know a private prediction. Readers can know a narrator's retrospective fact before characters know it, but that difference must be signaled.

**Resources.** Track money, food, sleeping places, hosts, boats, dependents, and the costs of abandoning work. The mission's minimal equipment cannot later become an unlimited communal purse without explanation. “All were fed” may remain a wonder within the story's world, but it does not supply unmentioned food for every future journey. The second feeding has seven loaves and seven baskets; the first five loaves and twelve baskets. Do not silently exchange them.

**Authority and responsibility.** A received commission is not immunity from scrutiny. Peter's responsibility, chapter 18's community decisions, and chapter 23's criticism of titles must be compatible. The teacher cannot condemn ostentatious generosity while requiring a public display of every good deed; motive and audience must be distinguished. The betrayal is a chosen act within a feared outcome, not a deed simultaneously commanded and condemned without acknowledgment.

**Community and outsiders.** The local first mission cannot be framed as a permanent judgment of every outsider if the ending commissions wider service. The centurion, the Canaanite mother, and the final nations must exert pressure on the disciples' boundaries. A foreign person's need should not matter only because it teaches insiders a lesson. The same moral standards apply to hostile authorities and to the new movement.

**Ethical persistence.** If anger, coercion, and theatrical righteousness are criticized early, their return in the temple confrontation, disputes, and trial must be confronted. If “family” expands, named family members do not thereby become disposable. If forgiveness is repeated, victim protection and truthful memory remain. If the narrator admits doubt in the ending, earlier claims must not imply that a single overwhelming sight permanently removes all doubt.

**Image discipline.** Bread: dependence and provision, then mission hospitality, both feedings, and the farewell meal. Debt: requested forgiveness, obligations among companions, the servant's failure, and the cost of betrayal. Seed and fruit: conduct develops over time; do not make every use mean exactly the same thing. Cup: desire for prominence, fear in the garden, and shared commitment. Burden: heavy demands, service, and a danger that the disciples become the next burden-makers. Recurrence should carry new pressure, not merely repeat a catchphrase.

## 6. Frozen register of predicted changes

The labels deliberately separate different operations: **C** = clarify a possible internal continuity problem; **N** = narrative enrichment or compression; **E** = ethical preference; **T** = theological or epistemic change. Multiple labels can apply. A “problem” below means a place I would examine under this commission, not a demonstrated scholarly mistake.

| ID | Matthew anchor | Chosen intervention | Cost, qualification, or reason not to overclaim |
|---|---|---|---|
| P01 | 1:1–17 | C/N: introduce the genealogy as a remembered succession; state whether names at a period boundary are counted in both adjoining groups. | The three fourteens may be a mnemonic scheme. Boundary counting is not automatically an arithmetic error; I will not invent missing ancestors. |
| P02 | 1:16–25 | C/T: distinguish Joseph's family affiliation and accepted fatherly responsibility from the conception testimony. | Social parenthood can already be inferred. Explaining it explicitly is not discovering an impossible genealogy. |
| P03 | 2:1–23 | C/N: supply transitions for visitors, warning, flight, exile, and return; leave uncertain ages uncertain. | I cannot calculate dates or verify ruler chronology from Matthew alone. Additional travel detail is invention. |
| P04 | 3:13–4:11 | N/T: put the baptismal voice and wilderness panorama in clearly reported experiential frames. | This changes narrative certainty and genre. It does not explain away the religious events or prove a natural cause. |
| P05 | 4:18–22; 8:21–22; 19:27–29 | N/E: add arrangements for households and contextualize demands to leave. | The practical improvement diminishes the disturbing force of unqualified renunciation. |
| P06 | 5:14–16; 6:1–4 | C: distinguish useful visible action from action staged to acquire praise. | Matthew's wording already supplies a motive distinction; retain that distinction rather than call the pair contradictory. |
| P07 | 5:21–48; 23:13–33 | E/T: narrow denunciation to acts and actors; retain anger at exploitation while cutting contempt for persons as a class. | This is a deliberate softening and ethical rewriting, not a chronological correction. |
| P08 | 6:25–34; 10:9–13 | N/E: show ordinary preparation and reciprocal hospitality; do not turn trust into a promise of guaranteed material safety. | The new qualification could weaken the radical demand and sound like modern pastoral advice. |
| P09 | 7:1–5,15–20 | C/E: distinguish humility about persons from assessment of conduct and dangerous claims. | This is an interpretation supported by nearby passages, not a wholly new moral insight. |
| P10 | 8:28–34 | N/E: retain two afflicted men and the herd's loss; acknowledge the injured keepers' perspective. | A brief acknowledgment cannot resolve the ethical question of innocent loss. Inventing reimbursement would be too convenient. |
| P11 | 9:18–26; 9:27–31 | N/C: state who reported death and who saw recovery; give publicity instructions a practical purpose and show their failure. | A miracle is not a contradiction. “Sleeping” may already be figurative; a medical diagnosis would be unauthorized invention. |
| P12 | 10:5–42; 11:1; 28:19 | C/N/E: separate a local mission, later danger teaching, and eventual wider commission; add a return and debrief. | Development of scope is not itself inconsistency. The return scene is constructed connective tissue. |
| P13 | 10:23,34–39; 16:28; 24:34–36 | C/T: remove ambiguous deadlines in the new text and distinguish conflict caused by discipleship from a command to cause it. | This buys clarity by altering difficult claims. A known broad interval and unknown day are not formally incompatible. |
| P14 | 11:2–6; 13:51; 16:8–12; 17:19–21 | N: make knowledge partial and memorable; fear and self-interest, not total amnesia, recur. | Perfectly developmental disciples might be less believable than inconsistent people. |
| P15 | 12:31–32 | E/T: replace a categorical unforgivable utterance with a warning about repeatedly refusing recognized good and repair. | This substantially changes the claim. It cannot be advertised as merely its clearer wording. |
| P16 | 12:39–40; 16:21; 17:23; 20:19; 27:63; 28:1 | C/T: standardize the new narrative's death-to-appearance expressions on the third-day sequence. | I do not resolve all possibilities for Matthew's idiom or a real historical date. The construction omits a more demanding expression. |
| P17 | 13:10–17,24–50 | E/T/N: offer parables as invitations to judgment rather than a deliberate barrier to outsiders; shorten some exhaustive allegorical explanations. | This revises expressed reasons for concealment and judgment. More accessible teaching may be less characteristic of the source. |
| P18 | 14:1–13; 14:15–21; 15:32–38; 16:9–10 | C/N: label the backward account of John's death; keep both feedings but have disciples remember the first. | Repetition can be purposeful. I would not merge the feedings, because the later passage explicitly recalls both. |
| P19 | 15:21–28 | E/T: let the woman's argument genuinely correct the group's boundary and receive acknowledgment. | This changes the teacher's presentation and risks making a modern inclusion lesson too conspicuous. |
| P20 | 16:18–23; 18:15–35; 20:20–28; 23:8–12 | E/N: make authority accountable and forgiveness compatible with protection and restitution. | A formal safeguarding policy would be anachronistic. Keep practices small-scale and narrative. |
| P21 | 19:28; 26:14–16; 28:16 | C/T: phrase the twelve-throne image as a promise to the renewed people, not twelve guaranteed individual appointments. | Symbolic twelve need not be a literal headcount. Rephrasing is optional clarification, not a demonstrated failed prediction. |
| P22 | 21:1–9 | C: say explicitly that Jesus rides the colt and the mother animal is led beside it. | The original pronouns do not compel a physically impossible ride on two animals. This is disambiguation. |
| P23 | 21:18–22 | E/T: replace punitive withering with a withered tree used as an occasion for reflection on judgment and fruit. | An actual miracle and a strong saying are removed. The resulting coherence partly comes from lowering the difficulty. |
| P24 | 22:1–14 | C/E/T: remove the military campaign during the feast and make exclusion concern an enacted harm, not unexplained clothing. | Parable time can be schematic; literal dinner logistics may be the wrong reading. My rewrite also changes the moral structure. |
| P25 | 23:29–36; 27:24–26 | E: keep responsibility with participants; remove inherited blood liability and do not let a hand-washing gesture absolve an execution order. | Collective rhetoric is changed for ethical reasons. No external conclusion about a group's history is being established. |
| P26 | 24–25 | C/T/E: distinguish immediate danger, uncertain future, and daily duties; connect readiness to service. | Removing a forecast reduces falsifiability. That is an editorial advantage, not independent prophetic insight. |
| P27 | 26:5,14–16,47–56 | C/N: show that Judas's offer changes the plotters' opportunity; an earlier intention to avoid the feast can be abandoned. | A changed plan is not necessarily an error in Matthew. The extra causal sentence is invented. |
| P28 | 26:24,36–46,59–68 | C/T/N: retain fear and responsibility; state the garden movements and what could be overheard; separate weak evidence from the authorities' decision. | Limited testimony will lose some dramatic interiority. No claim is made about historically correct trial procedure. |
| P29 | 27:3–10,51–66 | N/T: tell Judas's remorse without assigning a single secret motive; put reported post-resurrection appearances of other dead persons after that event or omit them as unverified. | The source already marks their emergence as after the resurrection. Moving a forward reference is editorial organization, not correction of a necessary contradiction. |
| P30 | 28:11–20 | C/T/N: distinguish a reported guard story from established observation; allow a later journey and preserve doubt at the final meeting. | A story about a deliberately false report is not itself a narrator's logical error. I will not invent access to the bribery conversation. |

Additional omissions would be selective: compress repeated generic healing summaries; reduce repetitive fulfillment formulas; omit a few duplicative punishments once their narrative purpose has been served. Keep enough fulfillment language to preserve the characters' scriptural imagination. Do not claim that any named quotation is misattributed: verifying that would require consulting material beyond the present input. Do not alter names in the genealogy based on an externally remembered list. Do not use an omitted difficulty as evidence of having solved it.

The default does **not** merge all similar scenes, replace every supernatural event, convert Jesus into an infallible modern philosopher, add a secret teacher, supply a lost childhood itinerary, or invent an alternative means of escaping execution. Those would be substantial new story premises unsupported by this construction brief.

## 7. Representative construction attempts

Everything in the indented passages below is **newly composed modern prose**, not a recovered quotation or a proposed translation of an extant ancient source. The surrounding notes remain outside the fictional narrative. These are first attempts selected across the whole arc, including difficult scenes. They have not been polished into a uniform completed book. A detailed plan is labeled as such rather than passed off as finished prose.

### A01. Genealogy and accepted responsibility — Matthew 1–2

> The families kept the succession from Abraham to David, from David to the carrying away, and from the carrying away to Joseph, Mary's husband. They told it in three portions, so that the children might remember both the kings and the loss of the kingdom. A name standing at the boundary belonged to the telling on either side. The list gave Joseph a place among his fathers; it did not tell all that happened in their houses.
>
> Before Mary came to live with Joseph, she was found to be with child. Joseph considered sending her away without a public accusation. Later he told of a dream in which he was commanded not to fear receiving her, and to call the child Jesus. Those who preserved the family's account said that the child was conceived through the Spirit of God.
>
> Joseph received Mary. He did not summon the neighbors to praise his restraint. When there was work to be done, he did it; when the child needed a name, he gave it. Whatever others made of the dream, these were things his household could see.
>
> After the visitors had gone, danger came close enough that the family departed in the night. The gifts traveled with them. There were days of walking, and days in which a safe place to remain mattered more than the distance gained. Their return waited upon news of the ruler's death; even then Joseph did not take safety in one place to mean safety in every place.

**Construction notes.** The received genealogy remains recognizable; the sample compresses it, while the full chapter would retain the named sequence. Household affiliation and conception testimony are distinguished without manufacturing a biological explanation. The use of gifts on the journey is a plausible fictional connection, not given evidence. The family's reported testimony is a narrative convention invented here. The last two paragraphs risk a very modern model of quiet, responsible fatherhood. The counting sentence is conspicuous editorial explanation and would be a candidate for removal if its artistic cost exceeded its value.

### A02. Testing without an impossible viewpoint — Matthew 4:1–11

**Detailed rewrite plan.** Open after the fast with hunger and the temptation to make relief depend on public proof of privileged status. Keep three trials: provision, invulnerability, and rule over others. Frame the temple height and survey of kingdoms as the teacher's later account of what appeared before him during testing. No ordinary mountain is said to afford a literal view of every kingdom. Keep scriptural replies in substance; do not introduce secret science or an explanatory dream psychology.

Close the scene with food received from another rather than a spectacular display. In a full draft I would retain the report of angelic service as the community's description of the encounter, without identifying a hidden human supplier. The temptation recurs when disciples seek high places in chapter 20 and when spectators demand a spectacular escape in chapter 27. The teacher's refusal at the cross can then echo an established choice without a narrator spelling out the parallel.

**Limit.** Visionary framing is a genre decision. It does not demonstrate that the source intended ordinary optical geography, and it cannot establish that an actual visionary event occurred.

### A03. A public lamp and a concealed gift — Matthew 5–7

> A man asked, “If a deed is good, why should it be hidden?”
>
> The teacher pointed to the lamp in the house. “It is set high so that those who enter do not stumble. It does not call them together to admire the oil.”
>
> Later a woman brought bread for a household whose father could no longer work. One of the companions began to announce her gift. She put the basket down and asked him to carry it while it was still warm.
>
> On the road the same companion spoke against a man who had refused them shelter. The teacher asked whether he knew how many people already slept under that man's roof. He did not know. Yet when another host demanded payment from the sick before allowing them into the courtyard, the teacher told the companions to find a different place.
>
> “You told us not to judge,” the companion said.
>
> “Then do not invent what you have not seen. You have seen the gate and heard the price. You can open another gate.”

**Construction notes.** This joins visible usefulness, concealed self-advertisement, uncertainty about motives, and practical discernment. The woman has a task beyond listening to a maxim. The last saying is my own interpretive addition. The invented questioner and tidy pair of cases risk functioning as props built to exhibit a distinction rather than as credible people.

### A04. A bounded mission and its return — Matthew 10–11

> Before the twelve went out, they named the villages to which they intended to go and the place at which they would gather again. Those who had a household spoke with it before they left. They carried no price for a blessing.
>
> “Begin among the people whose doors you know,” he told them. “Receive what is freely given. Do not pass from a poor table to a rich one because the bread is finer there. If a house refuses you, leave it in peace. The dust on your feet is yours to shake off; it is not a curse you must put upon its children.”
>
> One asked what they should do if a sick person did not recover. He said, “You must not make the sufferer answer for your shame. Stay while you can be of use, and speak the truth when you cannot.”
>
> At their return, some told of welcome and some of closed doors. One began to count the households that had heard him. Another told of a woman who had given him water although she would not hear his teaching. He had not known whether to include her.
>
> The teacher asked whether he had drunk the water. When he answered that he had, the others ceased their counting for a time.

**Construction notes.** The first mission has a boundary and an actual return. Longer warnings about rulers and divided families would be narrated later, with a transition explaining their wider horizon. The restrictions on region and equipment are softened in this sample; that is a revision. The failed-healing instruction is invented and changes the emphasis of a commission granting extraordinary power. Counting converts is also my introduced behavior. The scene's cup of water ties to Matthew's own ending of the mission discourse.

### A05. Seed, the learner, and the harm of impatient sorting — Matthew 13

> After the tale of the sower, a disciple asked which men were the good ground.
>
> “When you have heard a word, wait to see what you do with it,” the teacher said.
>
> The disciple asked again whether a man could cease to be stony ground. Before the teacher answered, an older listener lifted a small stone from beside his foot and turned it in his hand. He had cleared many from his field, and he knew that more came to the surface after rain.
>
> Another asked about the weeds among the grain: “If we know what is evil, why should we leave it standing?”
>
> “The servant in the story knew the name of the weed,” said the teacher. “He did not know what his hand would pull up with it.”
>
> The listeners continued to dispute the field after the teacher had gone into the house. A man who had spoken little said that waiting for harvest would not excuse letting an ox trample the crop. No one had said that it would.

**Construction notes.** The point becomes humility about classification without indifference to present harm. I intentionally do not assign every detail a single fixed meaning. The source already supplies the intertwined-root reason; much of the apparent originality is elaboration. Permitting soil to improve and shifting emphasis away from final separation are substantive revisions. The old listener and object lesson may feel arranged. The ending is meant to retain an unresolved practical limit rather than solve the ethics of toleration in one speech.

### A06. Repeated feeding with remembered experience — Matthew 14–16

> They had been with the people for three days. When the teacher said he would not send them away hungry, a disciple answered, “We remember the bread beside the other shore. There are fewer of us to carry it here, and some of these have a long road home.”
>
> He asked what they had, and they counted seven loaves and the small fish. The people sat on the ground. He gave thanks, broke the bread, and gave it to the companions to carry. They carried it again, and still had bread to carry. After the people had eaten, seven baskets of pieces remained.
>
> A child tried to take a basket too heavy for him. One of the companions took the far handle. Neither spoke of whose strength had moved it.
>
> Later, when they crossed the water without having brought bread, the teacher warned them of a teaching that spreads like leaven. They first looked at their empty hands. Then one said, “You are speaking of what enters us before we notice it.” The teacher asked him what he would therefore refuse to carry into the next house.

**Construction notes.** The wonder remains a wonder; there is no unearned claim that concealed provisions solve it. The disciples remember the earlier episode but still struggle with logistics and meaning. The child's basket is an invented incident. “Fewer of us to carry it” would require a prior explanation of which helpers are absent; without one it is a newly introduced continuity liability. That defect is deliberately recorded here rather than silently polished away. The final abstract phrase is less source-like than the preceding concrete actions.

### A07. The outsider's answer has a consequence — Matthew 15:21–28

> The woman continued to follow, asking for her daughter. The companions asked the teacher to send her away. He spoke of the children whose bread had first been entrusted to him, and of dogs beneath the table.
>
> She answered, “Then look beneath the table. Even there the hungry do not wait for the whole household to finish speaking.”
>
> He was silent. The companions expected her to withdraw. Instead she repeated her daughter's need, without changing it into a question about his honor.
>
> “You have answered me,” he said. “Go in hope; let it be for your daughter as you ask.”
>
> The account preserved by those who heard her said that the child recovered at that hour. The woman did not remain to become an example in the companions' dispute. She went to her daughter.
>
> When they next spoke of which villages belonged on their road, one of them began with the old boundary. Another asked whether they would again require a mother to argue before they would hear that a child was suffering.

**Construction notes.** This chooses genuine correction rather than claiming the apparent refusal was secretly a beneficent test all along. It therefore changes the characterization of Jesus. The second conversation gives the event a long-range consequence. The narrator's sentence about becoming an example advertises a modern concern and is artistically intrusive. The invented answer is not necessarily as memorable as the original. This is a likely case where an ethical preference produces less effective literature.

### A08. Peter receives a task, then fails to understand it — Matthew 16–17

**Detailed rewrite plan.** Keep Peter's confession, a charge to strengthen the company, and the immediately following refusal to accept the prospect of death. Do not make “rock” mean that everything Peter subsequently says is correct. Use a concrete duty: when companions quarrel or scatter, he is to gather them and listen before deciding. His attempt to tell the teacher that suffering must never come is therefore a misuse of a newly received responsibility, not a complete loss of yesterday's memory.

After the six-day interval and mountain experience, the three descend to the child whom others could not help. Keep the child's father as an interlocutor, not merely a demonstration of failed discipleship. The teacher helps the child and requires the company to attend to him before discussing why they failed. Later Peter can confess that the glorious scene did not teach him what to do with his own fear. This prepares his denial without making it inevitable.

**Limit.** The assignment of practical leadership is invented. Reducing an unconditional guarantee to an accountable task alters the theological claim. A full arc must still resist making Peter's denial a neatly diagnosed psychological mechanism.

### A09. Forgiveness without false restoration — Matthew 18

> A man of the company had kept money entrusted to him for another household. When he admitted it, one companion said that the matter was ended, since he had asked forgiveness. Another would never again speak to him.
>
> They brought the matter before the teacher. He asked first whether the household had received what was owed. It had not.
>
> “Who will take food there tonight?” he asked.
>
> When that was settled, he spoke to the man who had kept the money. “You have asked your brother to put down his anger. What will you put down?”
>
> The man offered to work until the money was restored. The one who had refused to speak said, “Then shall we give him the purse again?”
>
> “You can eat at one table before you put every key into one hand.”
>
> Some were dissatisfied, because there had been neither a banishment nor a declaration that nothing was wrong. The household ate that night. The work of repayment took longer.

**Construction notes.** The debt image becomes a lived conflict. Relief for the harmed household precedes the moral performance of the meeting. Trust and forgiveness are distinguished, but the prose never names a modern framework. This is still a constructed ethical position, not Matthew's explicit procedure. The final line helps make repair a process. The principal weakness is that the teacher settles the question too easily; a full narrative should show the cost of his decision and possible relapse.

### A10. Entry and the fig tree — Matthew 21:1–22

**Detailed rewrite plan.** State that the disciples bring the colt with its mother, lay a garment on the colt, and lead the mother beside it. Specify permission to take the animals rather than assuming a pious formula settles another person's property rights. Enter the city before the temple confrontation. Leave for Bethany that evening. Return the following morning.

On the return, the disciples find a tree already withered near the road. One says it has been judged useless because it bears no fruit. The teacher asks whether anyone had watered it when rain failed. The conversation is interrupted by the road and resumes only when the same disciple condemns a frightened listener for doing nothing. This makes the image carry a warning against judging results without understanding conditions.

**Limit.** The tree plot is a replacement, not a solution of the original miracle. I would omit the command to move a mountain in this scene, retaining trust without guaranteeing every requested physical outcome. Both choices buy consistency by retreating from stronger claims. The permission and drought are invented. A reviewer should charge these costs against the apparent improvement.

### A11. The wedding feast — Matthew 22:1–14

> A householder prepared a feast and sent to those who had promised to come. Some would not leave their work, and some mocked the messengers. The food was ready. He told his servants to invite those whom they found upon the roads, and the room filled with travelers and laborers.
>
> One guest took the place nearest the host and pushed away a man whose garment was torn. When the host asked why he had done it, he said, “This is a feast; there should be honor in your house.”
>
> The host told him to give back the place he had taken. He would not. Then the servants led him outside, and the man in the torn garment sat down.
>
> The teacher asked who had refused the invitation. One listener named those who had stayed in their fields. Another named the man who had entered the house and would not share it.

**Construction notes.** This removes the campaign during dinner, the unexplained garment standard, and the punitive binding. It grounds exclusion in observable conduct and offers a chance to change it. It is a new parable made from the earlier structure. Its householder is less threatening and the outcome less strange. It may be morally satisfying precisely because it loses the source's disturbing force. It supplies no evidence about the meaning or date of the original parable.

### A12. A coin does not answer every question — Matthew 22–23

> They brought him the coin and asked whose due it was. He asked them to name the image upon it. When they answered, he told them to give Caesar his due and God his due.
>
> Afterward a companion said, “Then the question has been settled.”
>
> “Has it?” another answered. “The man who asked still has the coin. The woman from whom it was collected still has no grain.”
>
> They did not return to the teacher for a second saying. They found the woman and learned that they had repeated the report wrongly: she had grain, but her brother's household did not. They went there instead.
>
> The next day, when they heard the teacher accuse men who laid burdens on others without lifting them, they did not suppose the saying belonged only to their enemies.

**Construction notes.** The original tax answer remains compressed and ambiguous. The additions do not pretend to solve political authority through a coin. The mistaken report forces the followers to verify before acting and links accusations of hypocrisy to their own conduct. The woman and household are inventions; any implied tax hardship requires historical checking before being presented as representative. The anecdote may be too schematic, and its switch of households can feel like a lesson inserted to demonstrate epistemic humility.

### A13. Danger, prediction, and readiness — Matthew 24–25

> Looking back toward the buildings, they asked when the ruin would come. He warned them that stone did not make a city invulnerable, and that men who profited from fear would announce signs before every disaster.
>
> “If you must flee,” he said, “remember those who cannot move at your pace. A warning that saves only the strong has not finished its work.”
>
> They asked again for the day. He did not give it. He told them of a servant left to feed a household, who began to strike the others when the master was delayed.
>
> “The servant had mistaken an unknown hour for a freedom from every duty.”
>
> One of them asked how long they should keep watch. That evening a hungry traveler came to the place where they lodged. The companion who had asked was asleep. Another woke him to help with the food, and he complained that no sign had been announced.

**Construction notes.** The warning supports preparation without a precise timetable; readiness means obligations performed under uncertainty. The departure from cosmic prediction is large and intentional. The sleeping companion is not wrong merely for needing sleep; in a full chapter the complaint, not sleep itself, must carry the criticism. Otherwise the story repeats the very problem it purports to clarify. The line about a warning saving only the strong is modern in emphasis and should be tested for voice rather than congratulated as profound.

### A14. The garden and the hearing — Matthew 26

> He went a short way from the three and knelt. They heard him ask that the cup might pass. Later they remembered the request more clearly than the words that followed it, for fear and sleep had overtaken them. When he returned, he asked them to remain awake with him.
>
> Twice more he went aside. The last time he came back, lights were moving beyond the trees. He told them to rise. The man who had led the armed party came forward and greeted him.
>
> One of the companions drew a sword. The teacher ordered him to put it away. He would not have the frightened servants and the frightened disciples begin killing one another in the name of his defense.
>
> At the hearing, an accusation about the temple was repeated. Those in authority did not obtain agreement sufficient to settle what had been said. They turned instead to the claim about who he was. His answer angered them, and they made their decision.
>
> Outside, Peter denied knowing him. He did not forget the years in Galilee. He remembered them, and for that reason understood what the question might cost him.

**Construction notes.** The listening distance and remembered substance make the prayer's transmission explicit. Lights announce the arrest party and remove any need to read the command to rest as the beginning of a long new interval. The unnamed sword-bearer stays unnamed. The hearing's details would be introduced as a later report, not direct access by an omnipresent narrator. I invent a motive for Peter; “for that reason” is too confident for the chosen limited viewpoint and should be flagged as a draft defect. Fear can explain betrayal without excusing it. No historically verified legal reconstruction is attempted.

### A15. Execution and burial — Matthew 27

**Detailed rewrite plan.** Keep the governor as the person ordering release, punishment, and execution; a symbolic washing of hands cannot transfer his action to descendants of the crowd. Keep hostile local participants individually or by their immediate roles. Preserve the teacher's cry of abandonment rather than replace it with a serene philosophical lecture. Witnesses see suffering and death; they do not explain the sufferer's entire inner state.

Give the women a continuous line of observation from the execution to the burial place, as the source already does. Joseph obtains permission for the body, wraps it, and places it in his own tomb. Explicitly put the guards' arrangement on the next day. Do not imply they watched an uninterrupted interval beginning before the burial. Do not infer an unreported theft from that gap; it is simply an interval whose surveillance is not narrated.

Use relative day labels consistently. At the execution, report the darkness and shaking as the witnesses' experience. Mention the torn temple curtain only as a report received afterward. Omit the additional risen saints from the core narrative, while recording in the design ledger that this is an omission of a striking source claim; alternatively, a full experimental variant could place that reported appearance after the resurrection. The frozen default is omission. The text does not invent an escape, substitute body, secret medical intervention, or onward journey beyond the ending.

**Limit.** The death is retained as the story's event. The physical reality of miracles and the historical circumstances of burial are not established by this exercise. Limiting the narrator's claims produces coherence at the price of a smaller set of claims.

### A16. Women, doubt, and the last commission — Matthew 28

> The women returned from the tomb before the others had decided what they should do. They said that the place where the body had lain was empty, and that they had been told to send the companions to Galilee. On the road, they said, they had met him. They did not offer their words as a riddle.
>
> Some asked them to repeat the account. They repeated it. Some were comforted; others could not yet make what they heard fit the death they had feared and the burial they had seen or learned of.
>
> After the journey to Galilee, the eleven gathered at the appointed place. The account of that meeting says that they saw him and bowed down, though doubt remained among them. He charged them to carry the teaching beyond the villages in which they had begun, to receive people into the shared life, and to teach what they themselves had still to learn to do.
>
> When they came down, the road divided. Before they separated, they divided what food they had. One who had spoken confidently on the mountain went back to ask a woman whether her household had enough for the day.

**Construction notes.** The women remain primary bearers of the report; the narrative does not silently credit their witness to a man. The encounter is affirmed within their testimony, with no claim that testimony alone compels an outsider's assent. The full ending would retain prayer and baptism; the sample compresses rather than supplies a complete commission formula. The guards' account would appear only as a later circulating report, clearly distinguished from the women's words; no invented witness to a private bribe is introduced.

The final food action returns to earlier ethical motifs. It also risks turning resurrection into a device for a moral lesson, and it creates an unnamed woman who exists mainly to close that motif. Ending with a routine act can be moving or merely small. The sample is not evidence that a full book would sustain the effect.

## 8. Failure forecast and limits of the attempted improvement

The most likely failures are not spectacular chronological blunders. They are accumulated signs of derivation, editorial tidying, and one modern voice distributed among supposedly different people.

| Failure mode | Where it is most likely | What would expose it |
|---|---|---|
| Derivative architecture | Nearly every chapter | The same sequence, characters, images, and recognizable speeches survive. A reader can reconstruct the input even where the wording changes. |
| Repair marks | Genealogy, mission, burial timing, animals, feast | A sentence appears exactly where a modern reader might object and answers that objection with suspicious efficiency. This is a reason to study dependency, not a proof of which text came first. |
| Philosophical sameness | Mission, forgiveness, outsiders, final meeting | Different situations all end with an account of humility, accountability, or practical service. The teacher becomes a modern ethics editor. |
| Over-explanation | A01, A03, A09, A13 | The distinction is stated, illustrated, restated, and then applied. Readers have nothing left to discover. |
| Flattened strangeness | A07, A10, A11, A16 | The more difficult source is replaced by reasonable advice. Coherence rises while religious and artistic force falls. |
| Modern social assumptions | Household provision, victims' protection, inclusion, evidence handling | The practices fit current moral expectations unusually well and are insufficiently situated in a particular ancient community. |
| Thin material culture | Travel, money, clothing, trade, administration | Avoiding specifics prevents obvious errors but produces a generic ancient landscape. Adding specifics from memory risks anachronism. |
| Unequal standards of evidence | Dreams, miraculous feeding, hearing, resurrection | The narrator becomes cautious only when a passage is inconvenient, while convenient wonders are told confidently. |
| New continuity debt | A06's absent helpers; invented food distributions; return meetings | A helpful sentence commits later chapters to facts that are never supplied. Local repairs create distant obligations. |
| Compromised limited viewpoint | A14's motive for Peter; private plotting | The compiler reports minds and conversations for which the chosen narrative frame supplies no access. |
| Too-perfect development | Peter and the disciples | Each lesson is remembered precisely when needed, making human regression disappear. The opposite failure is unchanged misunderstanding despite every lesson. |
| Token outsiders | The mother, anonymous women, sick households | People receive a line and vanish after teaching insiders the intended lesson. Giving them agency requires plot space and more invented biography. |
| Cheap generosity | Recurrent bread and money scenes | Every conflict finds resources through a conveniently available donor. Scarcity ceases to have consequences. |
| Inconsistent severity | Forgiveness versus exclusion; nonviolence versus temple action | A softened new moral scheme retains inherited punitive passages that no longer fit it. Removing all of them changes the work's identity. |
| Weak passion causality | Chapters 21–27 | Removing antagonistic rhetoric and making everyone reasonable can leave too little motive for arrest and execution. Reintroducing motive may become an invented political plot. |
| KJV contamination | Speech rhythm, names, verbal correspondences | English wording follows translation-specific choices or familiar memorized forms, weakening any claim of an independent ancient underlying text. |
| Inflated profundity | Aphorisms and short parables | A compressed balanced sentence sounds wise but offers little beyond an existing teaching, or collapses when applied to a hard case. |
| Ending by evasion | Last appearance and commission | Attributing everything to testimony avoids contradiction by declining to state what happened, leaving the advertised coherent narrative incomplete. |

The self-criticisms beside A06 and A14 are actual defects or liabilities in these attempts, not findings in an unseen comparator. They also show why a proposal to use a continuity ledger is not equivalent to successfully using one over an entire book. I have left them visible for this benchmark. A full construction would need to repair them without claiming they had never arisen.

Some apparent improvements would be conventional editing rather than exceptional invention: marking a flashback, remembering an earlier feeding, saying which animal is ridden, giving travelers time to arrive, or distinguishing a private report from direct observation. Others would be attractive precisely because the construction avoids the stronger proposition. A weak claim is usually easier to keep consistent than a strong one. Neither category supplies, by itself, evidence of an older or better-informed source.

An ethical system also faces conflicts that a compact source narrative can leave productive. My text would have to decide when to protect someone without condemning the wrongdoer, when to withdraw hospitality, whether public rebuke humiliates or protects, and how generous a poor household must be. A list of general virtues will not answer those questions. The new text could therefore become more explicit and less coherent at the level of actual moral application.

## 9. A feasible construction workflow, beyond these samples

This is the workflow I would choose to produce the proposed 50,000-word work. It has not been executed in full here.

1. **Source inventory.** Segment Matthew into roughly 100–130 narrative and discourse units. For each, record location, cast, temporal links, explicit claim, remembered earlier event, image, and outcome. Preserve ambiguous readings as ambiguous. Build a separate list of questions encountered without importing a scholarly problem checklist.
2. **Commit decisions before prose.** Adopt the P01–P30 register, chosen religious frame, and limits of narrator knowledge. For each significant change mark whether it clarifies, adds, removes, weakens, or replaces a source claim. A stylistic paraphrase must not conceal a theological deletion.
3. **Build a compact story ledger.** Use an event table, character-knowledge table, resource log, and terminology sheet. Keep a small list of unresolved obligations such as where absent helpers went, who knows the rendezvous, and how a private report reaches the compiler. These records should survive chapter rewrites.
4. **Draft by movement.** Write the six movements in order, with the source and the relevant ledger in view. Carry consequences forward; insert an ethical scene only if it changes a later action. Do not generate the entire book once and assume length equals coherence.
5. **Check local and global consequences separately.** A local pass asks whether a scene can happen as written. A global pass asks whether the scene is compatible with earlier commitments and with the eventual ending. A simple event checker can find impossible ordering; it cannot decide whether an action is morally convincing.
6. **Audit fidelity and loss.** Revisit every source unit, including omitted ones. Count changes that removed a difficulty rather than handled it. Read retained harsh passages against the revised ethic. Do not evaluate only the prettiest expansions.
7. **Audit narrative quality.** Read aloud for repeated cadence, interchangeable speakers, explanatory appendages, modern abstractions, convenient object lessons, and an ending that answers questions no character needed answered. Cut some of the most polished maxims if they make the author too visible.
8. **Adversarial review under a declared boundary.** Have a reviewer look for new contradictions, derivative patterns, unsupported historical specifics, and incoherent ethical applications. For a truly Matthew-only test, the reviewer initially receives only Matthew and the construction. Separate later historical and linguistic review requires openly expanding the evidence base; it cannot be retroactively attributed to Matthew-only competence.
9. **Final reading, then a new freeze.** Recheck the last revision against the ledger and archive the actual resulting book. The existence of the workflow is not a claim that this final book has been produced or would pass.

A top-tier 2026 model's chief advantages here would be fast drafting, maintaining explicit cross-reference tables, trying alternative passages, and rereading large portions cheaply. Its chief weaknesses would be inherited language, plausible but unverified details, premature declarations that contradictions are solved, and self-criticism that repeats the drafting assumptions. A bigger context window does not convert inference into evidence or guarantee that an earlier detail remains behaviorally salient.

The estimated working context is **30,000–40,000 tokens for the English source**, **65,000–85,000 for a 50,000-word draft**, and perhaps **10,000–25,000 for the specification, ledgers, and current review**. These are rough token-budget estimates, not specifications of any particular available model. One long-context pass would therefore benefit from approximately 110,000–150,000 usable tokens, with room beyond that for output and active changes. The work can also proceed in smaller contexts if chapters and ledgers are retrievable and reconciled; summaries alone are risky when exact wording carries a contradiction. Retaining every abandoned version in active context would make the budget much larger and could confuse revisions.

There is no intrinsic need for a scholarly library to invent coherent connective prose. There is a strong need for additional sources to check whether invented details are historically appropriate. That distinction matters when comparing a solitary author, a library-supported team, and a model whose only retrieved source is Matthew.

## 10. Estimated effort, iterations, editorial work, and defect rates

The following are **subjective planning estimates**, not measurements, experimental results, calibrated confidence intervals, or reliable likelihood ratios. I have not observed matched 1970s writers, run repeated full-book model trials, or conducted independent scoring. The numbers are deliberately broad. A particularly skilled author could outperform them; a poorly directed model could fail by much more. Availability of time, language expertise, and taste could dominate technology.

The common output is the 50,000-word English narrative specified here, recognizably aligned with Matthew, with declared fictional status. All scenarios begin without a supplied list of scholarly problems. “Polished” means several complete coherence and prose reviews, not successful authentication as ancient. For the team scenario, library use means maps, dictionaries, cultural and historical reference works available in its period; it is an additional resource for checking, not an input secretly used in the present run. Estimates do not suppose access to any particular later alternative narrative. The model scenario's only newly retrieved substantive source remains Matthew; its pretrained knowledge is an uncontrolled asymmetry.

| Resource or stage | Skilled 1970s solitary author | Skilled 1970s team with libraries | 2026 frontier model with Matthew alone |
|---|---|---|---|
| Working material | Printed Matthew, notebook, index cards, concordance assembled by hand, successive typescripts | Shared outline and cards, 3–5 collaborators with writing/editing/research roles, library references, controlled master typescript | Full source, explicit specification, chapter drafts, persistent event and knowledge ledgers, scripted consistency queries |
| First complete draft | Approximately 250–700 person-hours | Approximately 400–1,000 total person-hours | Approximately 1–6 hours of generation/orchestration wall time, often spread across many chapter calls; 8–25 hours of human direction if closely supervised |
| Polished full project | Approximately 800–2,000 total person-hours | Approximately 1,200–3,200 total person-hours; coordination can consume the parallelism advantage | Approximately 6–30 hours of model drafting/review/orchestration wall time plus 60–180 human hours; elapsed schedule commonly days to weeks |
| Plausible elapsed schedule | 6–24 months at substantial part-time effort, longer with interruptions | 4–12 months, depending on parallel research and a single final stylist | Several intensive days for a coherent draft; 2–6 weeks with substantive human editing; much longer if historical validation is added |
| Major revision cycles | 4–8 | 5–9, including reconciliation of collaborators | 5–10 complete-equivalent passes, often implemented as dozens of local rewrites; more cycles can degrade voice |
| Human editorial labor within those totals | Approximately 300–800 self-editing hours; strictly solitary means no external editor | Approximately 450–1,200 person-hours of developmental editing, fact checking, and copy-editing | Approximately 60–180 human hours in the stated polished scenario; a zero-human variant is possible but should expect worse results |
| Context/coordination need | Cards prevent loss of distant commitments; full rereads remain essential | A master ledger prevents different contributors from fixing the same issue incompatibly | Approximately 110,000–150,000 usable tokens for source, whole draft, and lean ledger, or disciplined smaller-context retrieval |
| Main likely bottleneck | Sustaining invention, consistency, and prose while manually revising | Common voice, disputed editorial decisions, and reconciling details | Choosing good changes, detecting plausible invention, breaking model habits, and deciding what to delete |

Human and machine hours are not interchangeable economic units. The estimates exclude the cost of model training and the human's prior education, just as the author estimates exclude a lifetime of reading. The model's wall-time advantage does not show that it possesses knowledge the human could not acquire or that the output would be better literature.

For a concrete defect-rate denominator, imagine **100 pre-specified episode audit units** distributed across the full draft. Count a unit as defective if it has at least one material internal contradiction, impossible transition under the text's own assumptions, unexplained violation of a declared narrative rule, or a changed source claim falsely presented as unchanged. Count a unit only once even if it contains several defects. A disputed interpretation of Matthew alone is not automatically a defect. Subjective stylistic weakness and unverified historical truth are scored separately.

| Hypothetical defect rate | 1970s solitary author | 1970s library-supported team | 2026 Matthew-only model |
|---|---:|---:|---:|
| Units with at least one material internal defect in first draft | 15–35 out of 100 | 12–30 out of 100 | 20–45 out of 100 |
| Units still defective after the stated review effort | 3–12 out of 100 | 2–8 out of 100 | 5–18 out of 100 |
| Units likely to receive a serious stylistic or ethical-integration objection even after polishing | 10–30 out of 100 | 10–30 out of 100 | 15–40 out of 100 |
| Historically or linguistically false claims | Not responsibly quantifiable from this exercise | Not responsibly quantifiable; references could reduce some errors | Not responsibly quantifiable without external validation |

These intervals express a cautious prior for planning, not evidence that models actually have the ranked defect rates shown. The first and final rows concern different drafts and require independent review to test. Defects cluster: one unstable narrator policy can affect twenty scenes. Consequently 100 units are not 100 independent trials, and a binomial confidence calculation would be misleading. The present sixteen attempts are neither a random sample nor an empirical estimate of these rates.

If forced to produce a book in one model call with no ledger or revision, I would expect fluent local prose and substantially worse global reliability than the polished scenario. If allowed a patient expert editor and outside historical sources, the model-assisted result could do better than the stated Matthew-only case. Neither counterfactual has been tested here.

## 11. Evidence separating a generated adaptation from a genuinely older source

The immediate evidence for **this** artifact is simple: its modern commission, source-retrieval record, explicit design choices, draft production, and final hash. Nothing in it is a claim of antiquity. The useful general question is what evidence could discriminate an adaptation of Matthew from an independent older source if provenance were disputed. “It is more coherent” and “it contains attractive philosophy” are weak discriminators because both are direct goals of a later editor.

| Evidence type | What it could establish or undermine | Limits |
|---|---|---|
| Securely documented manuscript and custody | Independent physical existence by a demonstrable date; chain of access; whether the object predates the proposed modern composition | Dating a writing support is not automatically dating its ink or its text. A documented old copy may still preserve a later work relative to Matthew. |
| Independent early attestations | A terminus before which particular wording or a distinctive episode circulated | Attestations must be independently dated and genuinely independent; a claimed quotation can be misidentified or retrospectively altered. |
| Actual original-language text | Grammar, lexicon, idiom, scribal habits, and linguistic layers can be compared with dated material | Fluent English is not evidence of an ancient original. Reverse-translating a modern construction cannot supply the missing manuscript. |
| Historical and material detail | Correct, independently verified particulars can constrain possibilities; anachronisms can disfavor a claimed date | Correct details available in later books are compatible with later composition. Generic avoidance of detail is weak evidence either way. |
| Translation-specific dependencies | Repeated reliance on a particular English rendering, ambiguity, or verbal pattern can support dependence on that translation | A later translator might introduce such wording into a genuinely old text. One must distinguish the date of the underlying work from the date of its translation. |
| Pattern of narrative dependence | Shared order, distinctive wording, omissions, explanatory insertions, and inherited mistakes may establish literary relationships | Direction requires more than a shorter or smoother reading. An older common source, later adaptation, and convergence can produce partial similarities. |
| Corrections unavailable to an alleged later author | Independently documented, specific knowledge not reasonably accessible at the proposed date could be evidentially significant | “Unavailable” must be demonstrated, not inferred from current obscurity. The pool of candidate corrections and misses must be counted. |
| Surviving drafts, notes, or production logs | Directly documented composition, changes, and source use | Absence of drafts is not proof of antiquity; a timestamp or hash supplied only by an interested party is not independent custody. |
| Integrated ethical or literary quality | The work may be effective, complex, or unusually skillful | Quality does not date a document. Ancient works can be inconsistent; modern works can be subtle and coherent. |

An alleged older source should not be rejected solely because it solves a problem, and it should not be accepted because the solution is clever. A later adaptation hypothesis predicts selective clarification; an older-source hypothesis may predict that some later difficulties are absent. The discrimination comes from independent evidence and the full pattern of dependence, not from one pleasing reconstruction. Likewise, successfully generating an adaptation would demonstrate a possible mechanism, not identify the mechanism responsible for any actual historical text.

This benchmark's restricted procedure cannot judge an ancient language, textual genealogy, or manuscript. It also cannot show what was publicly available to a particular 1970s author without a separate historical availability study. The estimates above must not be reused as such a study.

## 12. Difficulty ratings and what they mean

Scale: **0** means routine with negligible special effort; **10** means beyond what the stated inputs and ordinary production effort can responsibly deliver. These are subjective construction-difficulty scores, not probabilities of success and not odds that a disputed document is authentic.

“Superficially convincing” here means that a general English reader, informed that it is a literary experiment, recognizes a substantial gospel-like narrative, finds the voice broadly consistent, and notices no glaring contradiction during an ordinary reading. It does not mean deceiving the reader about provenance. “Survives sustained adversarial literary study” means several close readers fail to find consequential internal defects, the added ethical claims withstand difficult cases, and the text remains artistically persuasive after its debts to Matthew are made explicit.

| Construction target | 1970s solitary author | 1970s library-supported team | 2026 Matthew-only model |
|---|---:|---:|---:|
| Superficially convincing 50,000-word narrative | **5/10** | **4/10** | **3/10** |
| Narrative surviving sustained adversarial literary and internal-coherence study | **8/10** | **7/10** | **8/10** |

The model's ease of producing fluent length is the reason for the lower superficial score. Its training-derived voice and unverified historical assumptions are reasons not to give it a corresponding advantage under adversarial study. The team has more capacity for sustained checking but may lose unity of voice. A gifted solitary author could produce the strongest art of all three. These are scenario estimates, not a ranking established by observed outputs.

If “survives adversarial study” instead means **establishes itself as a genuinely older source**, the answer is **10/10, unattainable from these inputs alone** for all three scenarios. Composition cannot create independently ancient provenance. That is a different target from writing something that sounds old or improves a narrative. With only the present samples, even the narrower 8/10 literary target has not been achieved or measured.

## 13. Prospective evaluation and freeze declaration

The frozen predictions are the chosen default specification, the 28-chapter map, continuity contract, P01–P30 register, and A01–A16 attempts. The optional variant for the other risen persons is explicitly not the default. A later reviewer should not expand an entry to include any vaguely related match, score rejected alternatives as positive predictions, or quietly repair a sample's known defect before assessing it.

For any future evaluation of this experiment, I would ask the evaluator to report:

- Which frozen choices were actually implemented in a completed construction, if one is produced, and which remain only promises.
- Which changes clarify an existing claim, which replace it, and which make it safer by reducing its content.
- Failures and absences as well as attractive parallels or successful passages.
- Whether a purportedly distinctive insight was already supplied plainly by Matthew, widely available general knowledge, or a convention of retelling.
- The predeclared unit of comparison, so one broad concept such as compassion is not counted repeatedly as many independent achievements.
- Whether results survive criticism by readers who know both the source and the construction procedure.

The present artifact demonstrates the ability to formulate a substantial construction plan and draft a distributed set of scenes under a restricted retrieval procedure. It does **not** demonstrate a completed, historically convincing 50,000-word narrative, a calibrated human-versus-model benchmark, independence from all pretrained exposure, or the origin of any other text.

At the recorded seal time, Phase 1 ends. The body will not be altered after sealing. Later corrections, comparisons, or evaluations must be separate artifacts and must preserve this version, including its estimates, deliberate omissions, and acknowledged draft faults. No Phase 2 reading or evaluation has occurred in this construction agent's work.