Skip to content
Bookloft

The intelligence

It reads the whole book. And it tells you why.

This isn't a summariser under another name. It's a reading apparatus built inside a publishing house to do what a committee does: read in full, argue from several angles, leave a record, and hand the editor whatever it isn't sure about.

How it reads

01

How it reads

Five steps. None of them gets skipped when there's a rush.

  1. 01

    In full, not a sample

    The manuscript is read end to end, first page to last. Long books are split into batches so they can be covered, but nothing is discarded: no sampling, no truncation. A reader who only reads the first twenty pages isn't a reader.

  2. 02

    With its own genre's rubric

    A novel, a poetry collection, an essay and a picture book are not judged by the same yardstick. Each profile activates its own scoring families and its own lens, with a sub-score per criterion. An illustrated children's book is not measured against an adult novel's rubric.

  3. 03

    Twice, to know when it hesitates

    Every work is read twice over. If the two readings agree, the score holds; if they diverge, that variance is recorded and becomes a reason for a human to look. Knowing when it hesitates is worth as much as the score.

  4. 04

    With a committee of different eyes

    The same work is read from the craft, from the market, from the editor, from the reader, and from someone hunting for its cracks. Depending on genre, more eyes join: the bookseller, a prize jury, the teacher, the child who will hear it read aloud.

  5. 05

    And a synthesis that commits

    At the end there is a written judgement, with its recommendation, its strengths, what would need work and what it would take for the answer to be yes. Not a grid of numbers: a text you can argue with.

Every score, with the quote that holds it up

02

Every score, with the quote that holds it up

No score travels alone. This is a real record from the platform, with sample data.

Selection · AnalysisProposal: Editorial review

El cuaderno de las mareas

Marta Iribarne

Literary fiction · 62,400 words · read in full


Scores by dimension

Writing
8.4
Structure
7.6
Originality
8.1
Characters
7.9
Market
6.2

From the text

“The lighthouse had been dark for nine years and still she climbed it every night, just in case.”

Why it goes to review

High variance between the two readings on Market. The voice is sure of itself; the commercial fit is arguable.

When it isn't sure, it says so

03

When it isn't sure, it says so

The system discards nothing on its own. At any sign of doubt the work goes up to an editor's desk with the reason written down. These are the signals that raise it:

  • The two readings disagree
  • Confidence on some dimension is low
  • The work brushes one of the house's red lines
  • Something needs checking on ethical or legal grounds
  • A passage is flagged as possibly not original
  • The author's identity could not be confirmed
  • The manuscript looks incomplete
  • The two editors don't see it the same way

The judgement isn't generic: it's yours

04

The judgement isn't generic: it's yours

Each house writes its own manifesto, its own red lines and the weight it gives to each dimension. The system learns from the record of past decisions — what was taken, what was passed on, and why — and uses it as context on every new reading. Two houses reading the same manuscript don't get the same judgement, and that is exactly right.

What it has done so far

05

What it has done so far

Measured on the platform, not estimated.

5,152,226
words read
5,047
quotes pulled from the text
16.2
dimensions scored per manuscript
1/2
analyses that end up on an editor's desk

What it doesn't do

06

What it doesn't do

  • It doesn't decide. It proposes, and an editor signs the decision.
  • It discards nothing automatically. In doubt, it goes to human review.
  • It doesn't train models on your manuscripts or your authors'.
  • It doesn't cut a book off for its subject. Content is analysed and flagged; the policy is your house's.
  • It doesn't invent a score without evidence. If it can't quote the text, it doesn't score.

Fourteen days to test it with your own books.

Upload a manuscript and let it read. Import twenty books. Produce a royalty statement. That will tell you.

14 days · nothing is charged today · cancel whenever you like