Evidence · a real run

A validation run, published.

This is the output of one validation run in the shipped product, on a real manuscript, shown with what the author decided about five of them — including the one the checker got wrong and the one it was overruled on.

Scope

What was analysed.

Five chapters, 16,495 words, imported as one project and validated whole. The manuscript is a house title — the operator’s own unpublished novel — so no third party’s work is involved and nothing was anonymised: no name, place or line of prose was altered.

The book it is drawn from is longer than the five chapters analysed. That matters for reading the findings below: a contradiction spanning chapter four and chapter fourteen is not something this run could see, and at least one finding here is correct about the excerpt and wrong about the book.

Run summary
Manuscript five chapters, imported as one project and validated whole
Run 6b6070f2
Provenance a house title — the operator’s own unpublished novel; nothing anonymised

Findings

Five of them, worked through.

Two are quoted verbatim from the run output; three are summarised from it and marked as such on the card, because the run they came from is no longer retrievable through the API and a quotation nobody can check is not evidence. Each carries its severity, the agent that raised it, and what the author decided — a genuine contradiction, an intentional choice overruled, or a question left open. The second and third are the ones worth reading: a writing tool whose checker cannot be overruled is a worse tool, and a page that showed only the findings that were right would be advertising.

P1 · canon · quality-orchestrator · Dismissed — the citation is wrong

Cross-chapter contradiction: the origin and timing of the Mr. Hendricks disclosure conflicts between chapters one and two

“Ch1 establishes Samuel telling childhood-era anomalies (including ‘the neighbor who was there and then was not’) to Hannah as a kid, with her ‘listening all the way down.’ Ch2 has Hannah say Samuel only raised Hendricks once, as an adult, at a Thanksgiving years ago … Synthesis confirms this is not an isolated in-chapter slip but a contradiction that originates in Ch1’s setup.”

What the author did: dismissed it as stated. The contradiction is real; its cross-chapter framing is not. Every passage the finding places in chapter one — the telling, “the neighbor who was there and then was not”, the Thanksgiving — is in chapter two. So is the listening: “she listened all the way down” is real text, at chapter two, scene one, which is the placement error itself rather than a phrase the book lacks. The aggregating agent over-reached a correct per-chapter finding into a chapter attribution the manuscript does not support. The per-chapter version below gets it right, and that is the one that was accepted.

P2 · memory · memory-drift-auditor · A genuine contradiction

Drift in Samuel’s childhood memory examples, cited by Hannah against his own reflection

“In Scene 1, Samuel recalls telling Hannah as kids about ‘the shadows in their grandmother’s house,’ ‘the night he counted two moons,’ and ‘the neighbor who was there and then was not’ — establishing these as things he told her and she listened to. But in Scene 4, Hannah has no memory of Mr. Hendricks at all and treats the topic as entirely new/shocking (‘Mr. who?’), contradicting the earlier framing that she’d already heard about ‘the neighbor’ from him and listened ‘all the way down.’ This is a direct contradiction between an established shared-memory scene and the re-reference in dialogue.”

What the author did: accepted it. This is the same contradiction the dismissed finding above tried to describe, and this one places it where it actually is: scene one against scene four, inside a single chapter. The two together are the useful pair — the per-chapter auditor was right, and the aggregation step that tried to widen it introduced a citation the book does not support.

P2 · reader · reader-advocate · An intentional choice — overruled

Samuel’s reversal from resolved silence to blurting “Mr. Hendricks” reads as underdeveloped motivation

Paragraphs after Samuel decides “I’m not running interference” and successfully deflects Hannah’s questions about the survey, he cannot hold back the confession — “it was already moving, twenty-three years of not saying it”. The internal line justifies the reversal after the fact rather than preparing it.

Summarised from the run output, not quoted — no verbatim record of this finding was kept, so it is not presented as a quotation.

What the author did: overruled it. The abruptness is the point: the character has been not-saying it for twenty-three years, and a reversal that arrived on schedule would be a different, worse scene. A checker that could not be overruled here would be worse than no checker.

P3 · reader · reader-advocate · Uncertain — left open

Hannah’s mystery correspondent is teased and then dropped without payoff

Hannah mentions “one person who still answers my email … the only correspondent I’ve got who takes the questions seriously” — a clear hook. The scene moves on immediately, and Samuel’s internal reaction undercuts rather than builds the intrigue.

Summarised from the run output, not quoted — no verbatim record of this finding was kept, so it is not presented as a quotation.

What the author did: left it open. The correspondent pays off later in the book, outside the five chapters analysed — so the finding is correct about what it was given and wrong about the manuscript. That distinction is why the analysis scope is stated above.

P1 · reader · editorial-director · A genuine contradiction

The “notice anomaly, then rationalise it away” beat repeats across at least three chapters without escalation

Independent findings from several validators each flag the same structural habit in their own chapter. Aggregated, the repetition is a manuscript-wide pattern rather than three local pacing notes, and the recommendation is a developmental review of the beat as a pattern rather than per-chapter fixes.

Summarised from the run output, not quoted — no verbatim record of this finding was kept, so it is not presented as a quotation.

What the author did: accepted it. The aggregation is the useful part: three separate chapter-level notes would each have looked minor, and the same habit named once across the manuscript is actionable.

What changed

What Loreon did, and did not, change.

Loreon did not change the manuscript. It did not edit a line, merge a fix, or alter canon. Each of the five shown above is a proposal the author accepted, overruled or left open, and the prose on this page is the prose as written.

What it did do is read five chapters at once and report where they disagree with each other — see how canon is maintained for the mechanism, and what Loreon does not do for the boundaries this run does not cross.