The Phraseless Read
See what the draft is doing before you change it.
Paste or drop a scene, essay, or chapter. Phraseless ranks its nearest measured shelf authors, shows the exact evidence, identifies what to protect, and gives you one reversible change to test.
The default Read runs locally in your browser. It is not a rewrite, grade, or authorship test.
The useful transaction
Orientation, protection, experiment.
- 01
Find the literary neighborhood.
See where your piece sits on the live public-domain shelf - your five nearest authors, ranked, with the measured reasons and limits attached.
- 02
Protect what is carrying weight.
Phraseless points to the exact line or effect worth preserving before revision flattens it.
- 03
Test one change without committing.
Edit a scratch copy, recompute the cited measurements, compare it with the original, and undo it in one click.
Open only what you need
The short answer stays short. The proof goes deep.
The first screen carries the decision. The rest stays closed until you want the larger analysis, its boundaries, or another opinion.
Free Distinctiveness ProfileAvailable from 60 words; the shelf ranking begins at 180.
- Your five nearest shelf authors as a ranked, measured position.
- Up to three measured reasons, each attached to evidence in your piece.
- One line worth protecting when the evidence supports it.
- One author-connected craft move only when your passage independently qualifies.
Full evidence ReadAvailable from 800 eligible words across five paragraphs.
- A four-sentence lead mirror: what the receipts support, kept in proportion.
- One primary protection and one reversible revision experiment.
- Two or three supporting signals rather than a wall of equal-weight findings.
- Recurring patterns, local style shifts, and up to five Keep / Test / Cut calls.
- For professional writing, generic-drift findings use a named plain-language baseline.
- A full evidence atlas with passages, locations, measurements, rules, and caveats.
Across pieces and across modelsSeparate opt-ins; neither changes the local default Read.
- A local Writer Profile can accumulate twenty numeric style measurements across pieces. It stores no prose, excerpts, document hashes, provider output, or literary match.
- A separate model comparison lets you choose an eligible model, preview the bounded evidence that would leave the browser, and accept or reject each returned recommendation.
What is live - and what is not
Eight named authors power today’s results.
Literary V1 uses 40 complete works by Jane Austen, H. G. Wells, Edgar Allan Poe, Mark Twain, Edith Wharton, Willa Cather, Charles Dickens, and Robert Louis Stevenson.
What about the larger 120-work shelf?
A 40-author, 120-work expansion has been processed, but it is not used in product results yet. Processing a work does not approve a claim. The larger shelf stays out until its rights, calibration, editorial, and ranking-promotion checks are complete.
The writer keeps the last word
A measurement can describe movement. It cannot decide improvement.
Every consequential result names what was observed, offers a possible reading, proposes a reversible test, and lets you choose the original, the revision, neither, or unsure.