How it was tested before release
We do not publish manufactured testimonials. We publish what we tested.
This beta read releases without customer testimonials, because there are no customers yet. Rather than manufacture the appearance of one, this page states what was actually checked, where the record of it lives, and what it does not prove. Every number below is counted from a file, not estimated. Four kinds of truth are named here and kept apart: engineering verification — what the machinery was run against; live-read verification — whole manuscripts actually read; independent editorial validation — none yet, and we say so plainly; and native-language review — none yet either, and we say that too.
The run log
Every run of this system’s own machinery is written to an append-only log that has no delete path. Since 15 July 2026 that log has recorded 8,118 runs — the room’s whole run history, including the verified runs of the edition it has since retired.
Those runs break down as 2,910 diagnostics, 2,843 gate scans, 2,111 document builds and 254 reseals of the sealed package.
280 of those runs failed, and all 280 are still in the log — the record cannot be edited, so a failure stays where it happened.
Those are runs of the system’s machinery, not reads of anyone’s book — the log says so itself, and we would rather state the limit than let a number stand for something it is not.
Since 12 August 2026 an automatic check has re-run the room and this website twice a day. 53 pulse runs are on record and 46 of them are green — and when the check cannot see the room, it reports red rather than green.
Batteries, gates and pillars
- At its last certification, on 23 September 2026, when the site still published eleven editions, this website’s own test battery ran 2,154 checks — every promise, the price, the layout and every edition — and all 2,154 passed, none skipped. The battery has since been re-aimed at the English-only site, and its new total will be published here at the next certification.
- The system’s own diagnostic runs 67 checks across four layers and reports a score, a status and a gap list. It stands at 67 of 67, at level, gaps: none.
- A rubric gate scores every piece of code in the room against the same machine criteria — size, documentation, error handling, secrets, network access, personal data — and it is green.
- Six pillars have to be proved before the room can be sealed at all: that every shipped file is wired and fires, that the rubric is green, that the diagnostic is at level, that one edition of the room stands and no second edition sits beside it, that the delivery service answers, and that every claim the system makes has machinery behind it.
A pillar that cannot be checked in the moment is recorded as a skip, with the reason, and a skip is never counted as a pass. The seal refuses to write while any pillar is red.
What was proved about languages — and why today the read is English only
- Every full-length read this house has proven end to end — the counting of the chapters, the eight gates, every measurement and every delivered document, checked against each other — was on a manuscript written in English.
- The documents themselves were built in thirteen languages and audited one by one for their craft vocabulary and their decision words. That proved the documents can speak a language. It did not prove the read can measure a novel written in it.
- In September 2026 we tested the measuring itself against real novels in each language, and it did not hold everywhere: some scripts were counted wrongly, and chapters written in other languages were not found as chapters. Some of that is already fixed; none of it has yet been proven on a full-length read in that language.
- So the rule is simple. Today The Reader’s Verdict reads manuscripts written in English. Each native language opens only when a full-length novel written in it has been read end to end and every number in its documents agrees with its own pages. The translated editions of this website are paused until then, so no page promises a read we cannot yet give.
The honest limit: no native-language read has been proven yet. Native languages are coming soon, and we will say so here, language by language, the day each one passes.
Self-testing, and we say so plainly
Before it was offered to anyone, this system was run against the founder’s own full-length novels. That is self-testing, not independent validation, and we say so plainly.
- The read went through the founder’s own novel end to end — 54,492 words, prologue through chapter twelve, every page — and the delivered verdict is on file: the 1–4 band as it stood, an estimated potential in the 7–8 band, significant revision, with thirteen findings anchored to her own pages.
- A second complete live read went end to end through a 114,430-word novel of the founder's: twenty-one units, twenty-one read, the arithmetic closing to the last sixty-nine words of front matter, six rounds of audit, and 701 quoted spans tested — 676 exact, 25 declared as the room's own words, and not one word of the author's prose invented.
- Carrying a whole novel without thinning it was proved on a 93,178-word, twenty-nine-chapter manuscript of the founder’s, opened read-only: every chapter carried, none duplicated, none dropped, and not one word lost between the source and the blocks. The second half was carried as completely as the first — the place where a long read usually degrades.
- Showing the flag on the page was proved on her manuscript too: the quoted line chosen mechanically rather than curated, the repeated phrases counted from the text rather than estimated, and the finding that had no quotable line correctly rendering nothing at all rather than inventing a quote to fill the space. The finished document was reopened and audited — three highlights, all three verified word for word in her book.
The read told its own founder her manuscript needed significant revision, and the house kept the report. It did not flatter its author.
The manuscripts this beta read has been run against on the record are roughly 54,000 words and 114,430 words, both read end to end, and roughly 93,000 words carried mechanically with no evaluation formed. We take manuscripts up to 120,000 words, and the machinery has been tested to that size — tested as carrying, not as a complete read, and we cannot guarantee a complete beta read above 120,000 words. Three different kinds of number, kept apart on purpose: what has been read, what has been carried, and what we suggest you send.
Behind all of it are years of evaluation work on her own manuscripts, from which the instruments were built — fourteen of her titles carry evaluation records, running from December 2024 to July 2026.
The full record, in full
Every figure on this page, and the four sections this page no longer carries — the wiring and the claim register, the platforms compared, the failures we kept, and the four walls — with the file each one was counted from printed beside it.
Word file. Nothing in it is asked of you, and nothing is collected.
What this page is, and what it is not
None of this is a promise that every reader will agree with the findings of a beta read. It is a record of what was checked, what failed, and what was fixed before the doors opened.
Writing in another language? It’s coming.
Today The Reader’s Verdict reads manuscripts written in English. Native-language reads are coming soon, and each language opens only once it has been proven on real full-length novels written in it. If you would like to hear the moment yours opens, write to us: support@lovingmywords.com.