Chapter 42 · The Staff
When the Report Says Nothing
A clean report is the hardest thing a checker can say honestly, because the sentence people read into it, "your chapter is fine", is not the sentence it is entitled to. This chapter is about what Storskarvia's quiet line does and does not promise, and about the runs where Storskarvia refuses to say it at all.
The line itself
When a fact-check finishes with nothing to show, the Facts tab holds one sentence:
No issues found in 38 checked claims.
There is no card, no summary, and no score. The number beside the Facts tab name, which counts open findings, stays blank. Any notes that apply sit above the line, and they are the part worth reading.
What the number counts
The number is the claims this check judged, after the claim reader had finished with your draft and after the selection caps were applied. It is not the number of sentences you wrote, and it is not a measure of how much of your chapter was examined.
A claim here is a statement about the real world that your text asserts or implies: a date, a price, a brand, a piece of technology, an attribution, a period detail. Each one is rewritten as a self-contained sentence with the dates resolved, then checked against evidence gathered for it. Chapter 35 covers that machinery.
Three ways a claim ends up silent
A checked claim can reach the clean line by three different routes, and the report does not distinguish between them:
| Verdict | What it means | What you see |
|---|---|---|
| Supported | The evidence backs what your text says. | Nothing |
| Insufficient | The evidence retrieved does not settle the question either way. | Nothing |
| Contradicted or misleading, then overturned | The checker flagged it and the Skeptic successfully argued your manuscript was right. | Nothing |
The middle row is the one that surprises people. A claim the checker could not resolve is counted as checked and then disappears, because a checker that reported every unresolved question would hand you a list of its own ignorance rather than a list of your errors. The third row is the Skeptic's doing, and an overturned finding is discarded outright, reason included, so there is no hidden list to go looking for. Chapter 37 is about that trade.
The practical consequence: a run where four claims were flagged and all four were overturned looks exactly like a run where nothing was ever flagged. Both say the same sentence with the same number.
What was never a claim
Plenty of your draft is not eligible to be checked at all, and none of it is represented in the count:
- Plot events, invented characters, and subjective statements. The claim reader is told not to extract them. Your fictional bank robbery is not a claim about the world.
- Opinions. The reader labels these and they are dropped before any lookup happens.
- Claims past the caps. A single check judges at most 150 claims, spread across the whole document, and argues at most five passages per repeated claim. Whatever the caps dropped is stated in a note above the result.
- Claims phrased as corrections. When the claim reader fact-checks in advance and writes "ZIP codes had not yet been introduced in June 1962" instead of what your text actually asserts, that claim is thrown away, because judging a correction produces a supported verdict and loses your error. The run is retried once, and any that are still lost appear in a note.
The notes that change what clean means
Up to three notes can sit above the clean line, and they matter more here than they do above a list of findings:
- Weaker sourcing than usual. Retrieval was degraded. Either web search came back empty for a quarter or more of the claims, so those were checked against Wikipedia alone, or half or more could only be corroborated against a single site. A clean result from thin evidence is a weaker statement than a clean result from healthy evidence, and this note is how you tell the two apart.
- A scope note. For example, "Checked 150 of 400 claims found, spread across the whole document. Check its parts separately to cover every claim." The unchecked 250 are not in the count and were not examined. A clean line under this note is a clean line about a sample.
- An agent banner. When the check was run by your own AI agent rather than by Storskarvia's staff, the result carries a warning saying so. The judgment there is your agent's, and a clean result from it is a first pass.
If a clean result arrives with none of these notes, the run was healthy and the sentence means what it says.
When Storskarvia refuses to call it clean
"No issues found" is a claim about your manuscript, and Storskarvia will only make it when claims were genuinely read and genuinely checked. Three kinds of run would otherwise finish empty for reasons that have nothing to do with your writing, and all three are turned into failures instead:
| What happened | What you get |
|---|---|
| Claims were found, and evidence could not be retrieved for any of them | "…could not retrieve evidence for any of them, so nothing was actually checked. This is a search outage, not a clean result" |
| The only claims found were corrections, so none could be judged | "…could not phrase the claims it found in a checkable way, so nothing was actually checked" |
| No claims at all were picked out of 200 words or more | "…did not pick out any real-world claims in 4,120 words, which at this length usually means the claim reader failed rather than that there is nothing to check" |
Each appears in the Facts tab in the form "Check failed: …. Try again.", and each ends by telling you it has not been counted against your checks. That refund is real: the units are returned to whichever pocket they came from, your monthly allowance or your prepaid checks. See Chapter 41.
What to do about one: press Check Facts again. The claim reader is sampled, so a second run on the same text often behaves. A search outage is worth a few minutes' wait rather than an immediate retry.
A clean line with nothing in it
Below 200 words, a draft with no real-world claims in it is entirely believable, so the run finishes normally and says "No issues found in 0 checked claims." A page of dialogue, a paragraph of pure invention, or a scene that mentions nothing real all produce it.
Read that as "there was nothing here to check", not as "this is correct". If you expected claims in the passage and got a zero, check that the text you meant to check had reached the server before you pressed the button.
The continuity version
A continuity pass has its own tab and its own clean line: "No continuity conflicts found across 64 details." The number counts the details it gathered, from your manuscript and from your Story Bible, not the comparisons it made. A conflict needs one entity described two incompatible ways, so a long list of details with no disagreement in it is a clean result.
One note is specific to this tab. When the project has no Story Bible entries, the pass runs in the lighter mode where it names the subjects itself, and it says so: "Checked without a Story Bible, so Storskarvia inferred the characters and places itself." An empty result in that mode is a weaker statement than an empty result checked against declared values. Chapter 39 has the detail.
Why there is no praise
Nothing in the checking pipeline has a channel for approval. The stages emit findings or they emit nothing, and the judge is instructed never to praise and never to comment on writing quality. There is no setting that turns encouragement on.
A tool that told you a chapter was good would be guessing, and it would train you to trust the next thing it guessed. If you want a reaction rather than a verdict, that is what Part VII is for: share the document with a reader and let them tell you in their own words.
Updated 2026-09-18 · Full contents · Something wrong or missing?