The first thing that reads your book is not a reader

One week the book is finding people. The next week it isn't.

No email. No strike, no warning, no ban, nothing you could appeal even if you knew where to go. The numbers flatten, and every explanation available to you is about yourself: it wasn't good enough, the interest ran out, you missed the moment.

The actual event took about four milliseconds and had nothing to do with the quality of the writing.

A rejection is a judgement. Somebody read it and decided against it, and at least there's a person at the end of it who could in principle be argued with.

What happens to most difficult books is not that. Nobody read it. A system sorted it — into a category, onto a list, below a threshold — on the basis of words sitting near other words, and the sorting is invisible from where you're standing. You get the consequence and never the cause.

Being precise about this matters, because the wrong diagnosis leads somewhere bad. If you think you were judged, you conclude the work was weak. And the correction you make next time is to write something less like the thing you actually wanted to write.

Machines read your book before people do, and they read it without reading it.

Ad networks, payment processors, recommendation ranking, app stores, mail filters, social previews. All of them classify text at a volume that makes actual comprehension impossible, so they don't comprehend. They classify on signals: terms, adjacency, density, category. It's fast, it's cheap, it works well enough across millions of documents, and it is completely blind in exactly one way.

To that kind of reader, a memoir about getting clean and a page selling drugs are the same document. A war correspondent's account and the thing being reported are the same document. An investigation into an industry and an advertisement for it are closer to each other than either is to a recipe.

The consequence is exact and it's brutal. The chapter that makes your book worth publishing is usually the chapter that gets the whole book buried. The strongest thing in it is the liability.

And because none of this is announced, your only available theory is that you weren't good enough — so the correction gets made in the manuscript. That is how a classifier edits a book it never read.

Softening works. That's the problem with it. Take the edge off, the signals thin out, and the book travels further — as a different book, without the thing that justified writing it. You have solved the distribution problem by removing the reason for distribution.

The paywall is not a shield. This one is believed almost universally and it isn't true. Payment is a transaction, not a boundary. Paid text is still crawled, still classified, still excerpted, still screenshotted, still reported by somebody who bought it. Putting a price on a page changes who can read it. It does not change what reads it.

Which leaves a third option most writers have never been offered, because most platforms have one switch and it says public or paid.

Public, sensitive, paid, private are not settings. They are four editions of access, the way hardback and paperback are two editions of paper.

That isn't a euphemism. It's the accurate description, and it changes what kind of decision this is. An edition is a publishing decision: the same book, produced into the world in a particular form, to reach a particular reader in a particular way. Nobody has ever thought a paperback was a censored hardback.

Tiering is placement, and placement is the opposite of softening. Softening changes the book so it can go anywhere. Tiering leaves the book alone and changes where it goes. Everything strong in it stays strong. It just stops being handed to the one reader that can't read.

That sentence reads badly on its own, so here is why it's the right way round.

Placing your own work means guessing at systems you cannot see, from the position of the person least able to be objective about it. You know what you meant. The classifier doesn't, and never will, and the gap between those two things is exactly where books get buried. Somebody who reads a lot of difficult work, and who is thinking about machines rather than about morals, will place it better than you will. Not because your judgement is poor. Because the question isn't really about your book — it's about the readers who aren't people.

Two things follow, and both are already true here.

Paid is not a shield. It doesn't become one by being a tier.

And the tier can move after launch. Classification isn't fixed: lists change, thresholds change, a book that was fine on Tuesday is not fine on Thursday. Placement is maintenance, not a decision you make once.

$30. Thirty minutes of our team's time — the work read, the tier set, unlocks priced, the page live at the right level.

You can do all of it yourself, here, for nothing. The tiers are visible and the controls are yours.

What the $30 buys is a reading of your book by somebody thinking about the machines instead of about you. That's the one perspective you cannot get from inside your own work, and it's the one that decides whether the sharpest thing you have written gets read or gets quietly filed.

Send the file.