My beta reader wrote to me just before leaving for class. They’d finished the first twenty chapters of Solar Gate, and on the way out it hit them what was bothering them, which in their words (my translation from Polish) was a plot hole stretching from Earth to the Hyades.

Some context first. Humanity’s biggest colonies are in the Hyades, a star cluster about 150 light-years from Earth, and for a century a wormhole connected them to the Solar System (Sol, in the book), until it collapsed eight years ago. Sol has spent seven of those years in a civil war between Earth, Mars, a loose federation of space stations and a crowd of smaller factions profiting from the mess, so medicine runs short and fuel is rationed by formula.

Then my protagonist, Tom Nowicki, arrives from the Hyades in a ship nobody in Sol can match. He wants the segments of an 800-meter ring from Earth’s shipyard, plus heavy tugs and fuel from Mars. From the stations he asks for forty years of survey records and the crews who used to work around the old wormhole, though he won’t say what any of it is for.

And they all say yes.

Who in their right mind, my reader asked, commits scarce resources in a war economy to something undefined, with no guarantees? Nobody negotiates, and every decision gets made by one person, as if each faction were a dictatorship. My reader didn’t buy it and doubted other readers would, so they suggested a chapter or two from the Solar side, where we’d finally see the factions’ point of view.

Feedback like this is how you avoid the 3-star reviews later.

Why a chapter in the middle is hard

I had a prologue and 24 chapters, about 58,000 words, with the rest of the book still to write. A chapter dropped into the middle has to agree with everything around it: who knows what by which chapter, which ship is where, what each character promised and to whom, even which month it is. Get one fact wrong and you’ve opened a new plot hole while closing the old one.

I’ve written this book in Claude Code from the start: I write the chapters, and Claude keeps them as markdown files next to the worldbuilding codex. The project folder holds the book’s whole paper trail, with a record of what changed in every chapter, a file of canon rulings, a running continuity sheet with every promise and open question, and my style guide. That paper trail is the reason the rest of this worked.

What happened first

Before I touched the manuscript, I pasted the feedback in and had Claude read the chapter records for chapters 9 through 24, along with the scenes where Tom makes his requests. It came back with a diagnosis that partly disagreed with my reader.

My reader wanted pushback, and the pushback is already in the book. In chapter 21 Earth holds back 62 of the 64 segments until Tom says what they’re for, while in chapter 22 Mars won’t release a loaded tug until one of its engineers can inspect the site. By chapter 24 all three factions have pulled out, but my reader stopped at chapter 20, right at the top of what my own notes call a false victory.

On the larger point, Claude was just as plain that my reader was right. In chapters 17 to 19 each leader’s reason for saying yes lived entirely in subtext, because Tom’s point of view never spells it out. Anyone in Sol would read a giant ring built where the old wormhole used to open as a possible way back to the Hyades. Saying yes to the only faster-than-light ship in the system also keeps it away from your enemies, and so far every commitment was small enough to withdraw, yet none of that was on the page. Worse, three leaders in a row declined to ask what it was all for, and three in a row looks like author convenience.

It also suggested that Earth’s leader should guess what the ring is and keep quiet about it, which plants something I can pay off in chapters I haven’t written yet.

Where the chapter came from

The plan I went with was a new chapter between 19 and 20, plus fixes in three earlier chapters, including one scene where a leader finally presses Tom hard on what he’s building.

I told the Earth scenes from the side of Lena Vasquez, Earth’s leader, so the reader would hear her reasoning from her. I gave Mars its own scene, with Adaora Obi, who governs Mars, taking the request to her board and nearly losing, since my reader had asked why nobody inside the factions ever pushes back. Both scenes had to show why each leader says yes. I put Earth’s cabinet meeting on the orbital shipyard, with the segments being cut outside the window, and made the head of Earth’s navy very British, in name and in the way he talks. Outlining each scene, even if only in my head, makes the writing much easier later.

Before I picked a name for him, I had Claude search the earlier books. “Pemberton” already belonged to an admiral from an earlier book, so he became Admiral Sir Giles Hollingworth, a man who says things like “I did think I’d misread that.”

(Every saga eventually has a name-collision problem, and several books in I no longer remember every minor officer I’ve named, so an assistant that checks the back catalog before I name someone saves me the embarrassment I’d otherwise discover in a reader email.)

The chapter

The new chapter (19a, until I renumber everything) has two halves. On Mars, Obi asks the board that shares out Mars’s resources to vote fuel and tugs for Tom with the purpose field left blank, and the board fights her. The member from the fuel office points out that Earth pays for that fuel while Tom pays nothing, and the member for the people on the planet’s surface says those construction crews were promised to a new hospital wing. Obi wins five to four, her narrowest vote ever, and only because the oldest member makes it conditional on an inspection before the fourth tug sails.

On Earth, Vasquez works out alone that the ring is a gate, a doorway back to the Hyades, and keeps it to herself. When she has to defend her yes to her cabinet, her answer to the admiral is the one my beta reader was missing: a price turns it into a contract, which a man with no authority can’t sign; ask him for one and he takes the same question to Mars. Hollingworth gets his price anyway, a frigate escort for anything that goes near Tom’s ship.

The chapters after it read differently now. In chapter 21 the frigate used to arrive as a surprise, and now it’s the admiral’s price, while the inspection Mars demands in chapter 22, which used to look like Obi changing her mind, is the price her board set. Events that already happened later in the book read as consequences.

The question that found the errors

When I finished the chapter, I asked one question: does it fit everything that happened before and after?

Claude went back through chapters 12 to 24, checking every fact the new chapter touched, and found one real contradiction. In my draft, the Mars board’s condition applied before a second tug sailed, but in chapter 21 three tugs arrive with no inspection, and the tug Mars holds back in chapter 22 is the fourth. So the vote became a first batch of three tugs, which also matches the cargo figures announced in chapter 21.

Then came the small slips I probably wouldn’t have caught myself. Hollingworth set down a hat that no line had ever given him, because an earlier edit had cut it. One woman was called the last member of the original Mars board still serving, though Obi had sat on that board too. A clock time was in ship time where it should have been Earth time. The Mars scene, which takes place before the end of chapter 19, had no time cue to tell the reader so, and a line in chapter 18 now contradicted the new vote.

Besides the new chapter, I made small edits to chapters 17, 18 and 19 to fix all of it, and the continuity notes were updated to match.

Read on its own, the new chapter looked clean, yet a separate pass against the surrounding chapters turned up six errors, and that pass happened because I asked for it.

The chapter has gone to my beta reader. Now I wait to hear what they think.

I wanted control over my books, and I got admin.

That is the quiet bargain behind indie publishing: you choose the story, the cover, the release date and the price, then discover that each choice creates another small job. The novelist gradually acquires a second occupation as the manager of a tiny publishing company.

I run mine across multiple pen names, and on a normal week I might work on a manuscript, check an Amazon ad, plan a newsletter, maintain a website, update launch tasks and inspect sales. None of these jobs is especially difficult, but together they can eat the week.

The switching is worse than the work.

You open a manuscript, remember that an ad needs checking, open the advertising dashboard and notice that sales were odd yesterday, and now you are in KDP; twenty minutes later, you are comparing two dates in a spreadsheet while the unfinished chapter waits in another window.

By lunch, you have been busy for hours and written nothing.

Most conversations about AI concentrate on generating prose, the one part of the job most writers want to keep for themselves. Sales reports, newsletters, ad checks, metadata and half-remembered experiments get far less attention.

I don’t need a machine to have my ideas for me, but I would happily give it the Tuesday afternoon spent working out whether a campaign became expensive or sales merely had a bad day.

A chatbot still needs a manager

An author can already hand a KDP export to ChatGPT and ask for an analysis, and the same chatbot can draft a newsletter or suggest ad copy. All of that is useful, but the author still has to remember what needs to happen, gather the files, explain the context and ask the right question, while the tool waits for someone to give it work.

That saves labour inside a task, but it does little about the mental burden of running the business.

For the past few weeks I have been running something different: AI employees with defined jobs. Plural, because I am testing two of them on the same work, Viktor and Grok Bot, and building a third of my own. Their responsibility is the marketing and admin around the books, while anything that touches my money, my readers, my reputation or my published work still comes back to me.

The distinction is simple: a chatbot says, “What do you need?”, while an AI employee says, “I noticed this, and here is what I think deserves your attention.”

What I no longer do myself

Ads first, because that is where an author’s attention leaks fastest. Every Monday one of my agents pulls Amazon and Meta performance and compares the last seven days with the seven before. It can read the numbers, but it cannot change a bid or a budget, so the worst outcome of a bad analysis is a bad paragraph in my Monday report.

Sales come from KDP exports. Because KDP has no public sales API, I still download the file myself, which keeps that step stubbornly manual. After that, both agents read the export knowing my series order and last week’s ad numbers, so I no longer explain my catalogue every time I want an answer.

Newsletters slip when the manuscript is going well, which is exactly when an author wants to stay inside the book. Viktor checks whether mine went out on schedule and helps me prepare the issues themselves, while I decide what goes in and when it is sent.

Content research runs twice a week for The Indie Lab: a pack of five to eight angles with a shortlist on top, and a rule against bringing back an idea I have already shortlisted unless something new has happened. The agent drafts no article and publishes nothing.

The last routine has nothing to do with marketing, and it exists to protect the reason for all the others. On Fridays one agent compares my calendar with what actually happened to the manuscript and tells me when a call or a meeting fell inside a writing block. Twice a week it looks at the folder for the book I am drafting and stays silent unless a chapter changed, or a writing session passed with nothing to show for it.

I don’t want an autonomous publisher

There is an obvious way for this idea to become stupid: give an AI access to advertising accounts, email tools, the author website and publishing platforms, then let it change budgets or contact readers without supervision. It would remove work in the same sense that a kitchen fire removes the need to cook dinner.

The sequence that works is more restrained: observe, analyse, recommend, prepare and ask for approval. In my setup the ad data comes in read-only and drafts stay drafts. Nothing goes out, and no money moves, until I say so.

The other requirement is memory. My agents know my pen names, the series order, the experiments I have already run and the rules I have set, because if I had to explain my catalogue from scratch every Monday, I would have acquired a needy intern with an excellent vocabulary.

It still needs supervision. Publishing data does not always arrive neatly, the KDP export is still my job, and one of my agents talks to me through Slack, which is a faintly ridiculous thing to install when the whole company is one person and a cat. Last month one agent’s scheduled checks went quiet for weeks, and I only noticed when I asked it what it had been doing. Managing the AI is a small job, but it is a job.

The big labs have noticed

At the end of September OpenAI launched Dots, always-on agents inside ChatGPT that keep working between conversations and come back when a decision needs a human. Grok Bot, one of the two I am testing, comes from xAI. The idea behind this post is now a shipping product from the biggest AI companies. European authors get a catch with it: Dots on ChatGPT Pro is not available to subscribers in Europe, so from Poland I cannot test it on a personal plan.

The business created by the books

I am not trying to escape the commercial side of publishing, because an indie author who ignores sales, costs, cash flow and reader acquisition is running a hobby with unusually elaborate paperwork.

I want my attention where it changes the outcome. Choosing what to publish, setting release timing, deciding how much money I can risk and understanding why readers continue through a series are my decisions. Copying numbers between systems and remembering that the newsletter is due are jobs I would hand to staff, if I had any.

Most authors don’t. The awkward middle group has enough books and revenue to create real operational work, but not enough money for an assistant and an advertising specialist, which is where an AI employee earns its keep.

Nobody can promise that an AI employee will sell a book, and anyone who does has probably found a more reliable business in selling promises to writers. My agents hand me something smaller: what happened, what needs a decision, what can be ignored and what can wait, so the marketing keeps moving while I work on the next book.

I set up Viktor for clients, so if you want to test it, send me a message.

My style guide for the Solar Gate series says there is no faster-than-light communication in this universe, with a single exception. In chapter 5 of Echoes of the Storm, the sixth book in the saga, a ship uses a bubble gate to hold a real-time conversation across twelve light-years. Both are mine. I never noticed.

Claude Code noticed, while checking the new book against the old ones. It’s a programming tool, and I now use it for more of my publishing work than anything built for writers.

Solar Gate is the seventh main book in the saga. The six main books and four novellas before it come to roughly 455,000 words, all of it canon now: ship names, faction politics, how many Star Cores exist and who holds them, which character calls another by his first name. Before going further with the new book, I wanted to know whether its prologue and first twelve chapters (about 31,000 words) still agreed with all of that. Nobody holds 455,000 words in their head, though series readers notice the slips anyway.

What else it found

The first continuity pass covered the prologue, the twelve chapters and my planning documents: codex, outline, concept doc, style guide. It came back with 23 issues.

The physics was wrong twice. Chapter 6 had sunlight arriving at Neptune “eight hours late,” when the real figure is about four. Chapter 7 described a signal from Earth as weeks old, though radio covers that distance in about four hours. The first error is off by a factor of two, the second by a factor of a hundred or more. Science-fiction readers write reviews about exactly this.

The dates had drifted. Solar Gate is set eight years after the previous book, when the crew has been together about ten. Chapters 1 to 11 held eighteen stale time references, most of them “five years” with a few “four years,” and the concept doc, style guide and outline said five as well. A blind search-and-replace was out of the question because “four years” was wrong in some lines and right in others. Claude listed each instance with the sentence around it, which let me check them one at a time.

The wormhole collapse needed untangling. That event, which cuts the Hyades off from Earth, is the premise of the whole book. My codex describes a scientist fleeing through the wormhole with a weapon, while chapter 4 has a prototype forced through the terminus, and chapter 1 talks about competing technologies tearing at the same patch of space-time. Side by side, they read like different events. Claude found the original scene in chapter 17 of Echoes, which let me read all of them against the source: they turned out to be one event seen from different angles. The line that was actually wrong sat in chapter 2, where “since the Hetman lost the wormhole” makes it sound as if the ship caused the collapse.

Then the small things. Chapters 9 and 10 used one set of faction names, while chapter 12 used another. One character was narrated as “Dmitri” in five chapters and as “Volkov” in the rest. This is what beta readers circle.

Why I left the writing apps

I tried the tools built for this. They were useful, but also rigid, because each one had its own idea of how an author works with AI, with fixed panels for characters and a fixed way to feed context to the model. My way of working didn’t fit. The configuration piled up faster than I could track it: I’d change a setting for one chapter, forget it, then wonder three weeks later why a session was suddenly behaving differently.

Claude Code is Anthropic’s coding agent, built to work across a large project: search it, read what’s relevant, change one file without breaking another. It runs in the Claude desktop app, pointed at a folder on your computer. Mine is a Google Drive folder that syncs to my laptop.

Why plain markdown

Markdown is plain text with a few symbols for formatting, such as a pound sign for a heading or asterisks for italics, which is about all a novel needs.

The files belong to me, in the plainest sense: they open in Notepad or any other editor on any machine, and they’ll still open in twenty years. There’s no export step, so nothing gets lost in conversion. If I want to try a different AI tool tomorrow, I point it at the same folder. Search runs across the whole saga at once, which is how a line in book eight turns up against chapter 4 of book eleven.

Drive keeps earlier versions of each file, with a catch that matters here. Google deletes old versions after 30 days, or once a file has 100 newer ones, unless you mark a version to keep forever. An agent that saves often can burn through 100 versions quickly, so mark the version you care about before a big revision. When the book is done, markdown converts to the Word file or ebook KDP expects.

The setup

Chapters/    one file per chapter
Codex/       characters, factions, ships, tech, timeline (one file each)
Notes/       running continuity state, chapter records, open contradictions
Reference/   style guide, outline, concept doc, and the ten earlier books, split one chapter per file
Editing/     the prompts for chapter-level and whole-book passes

The earlier books sit in Reference/ one chapter per file, so Claude can search them for a name or an event and read only the passage it needs. A short START-HERE.md in Notes/ tells every new session what to read and in what order. A file called canon-flags.md holds every contradiction found, each one waiting for me to rule on it.

Claude flags a problem. I decide what’s true. The ruling goes into that file, where every later session reads it before touching anything. The configuration is a text file I can open and read, so there’s no settings screen to lose track of.

Where the writing happens

I don’t use a separate editor, because the Claude desktop app shows the whole project in a file pane on the left, with the chapters, codex, notes and ten earlier books in one tree. Clicking a chapter opens it beside the conversation, where a switch to the source view lets me edit it in the same window Claude is working in.

Starting a chapter takes one sentence: I ask Claude to create 17_Chapter_17.md in Chapters/, then open the empty file and write. When a passage needs a second look, I select it and either attach it to the conversation or leave a comment on it, so the question stays tied to the exact lines. My old tools spread this over several panels; here it’s a file tree next to a chat.

Logging now, fixing later

When a book is moving forward, I don’t want to stop and repair chapter 6. The rule written into the notes is that contradictions get logged and ruled on, while the cleanup waits for a dedicated editing pass. Nothing forces a fix in the moment.

Anyone who has run a software project knows the pattern. Specs drift from code. Docs drift from both. By release, nobody’s sure which one is true. You don’t fix it by rereading everything; you keep one file that says what’s true and check the rest against it.

Where it’s worse

It costs money: Claude Code needs a paid plan. I pay $20 a month, on top of whatever else is in my publishing stack.

It’s a technical tool, since you work in a desktop app pointed at a folder on your own machine, not a website where you log in and start typing. It has no character cards or timeline to drag around, only folders and text files, which for some authors ends the conversation.

It’s also a general tool. It doesn’t know what a novel is until your style guide and notes tell it, so an author who skips that step gets generic results. My first session went on working out the file layout, which Claude helped with, though it won’t hand you a finished template.

And I have no sales numbers for any of this. Solar Gate isn’t out yet.

Who should try this

If you write standalones and your current tool doesn’t get in your way, I wouldn’t bother, because the payoff grows with the amount of canon you have to protect. In a long series, or a backlist readers binge in order, one contradiction in book eleven earns a one-star review from someone who read books one to ten last month.

Simon & Schuster launched Simon Stream in July, and the pitch is aimed straight at us. You don’t need an agent, submissions are open, they promise an answer within six weeks, and they pay 50% ebook royalties monthly, with Big 5 editing and marketing on top.

Before you open the submission form, look at what the headline signing did.

Jack Gatland is the imprint’s flagship author, and he’s Tony Lee, a #1 New York Times bestseller with close to half a billion Kindle Unlimited page reads. What he gave Simon Stream is a brand-new five-book cozy series, The Moriarty & Co Mysteries, out on 27 October. His existing series stayed where they were, and he keeps releasing new indie books next to it. A writer with decades in both camps licensed a side project to a publisher, and the catalogue that built his name stays in his own hands. Smart move, and not the story the press release tells.

50% of what?

Simon Stream pays 50% of net receipts on ebooks, where traditional publishers usually pay around 25%, so by trad standards this is generous, and S&S will tell you so.

Now put it next to KDP, because that’s your real alternative. Price an ebook at $4.99 and Amazon pays you about $3.49 a sale at the 70% rate, minus a small delivery fee. Put the same book through Simon Stream and Amazon pays S&S the $3.49, then S&S pays you half, so you get roughly $1.75.

There’s no advance to cushion that, so if S&S sells twice as many copies as you would have, you break even, and you’ve paid for editing and marketing with half your royalty for as long as the contract runs. If they sell fewer, you’ve lost money on a book you already wrote.

Doubling an established indie’s sales is possible, but it’s also a big ask of an imprint whose first flagship title hasn’t come out yet, so nobody has a sales record to judge it by.

What the FAQ doesn’t say

The FAQ is long and friendly, and it skips the clauses that decide whether a deal is any good.

How long do they keep the rights, and when do they come back? The page doesn’t say, and instead it offers to “discuss rights by format and territory,” which is where a contract lives or dies.

What do print and audio pay? Only the ebook rate is published. Every book gets a human-narrated audiobook, which is great, but it also means your audio rights sit in the deal at a rate nobody has disclosed.

Does the book stay in Kindle Unlimited? Books go to subscription channels “where rights allow,” and if your income is KU page reads, get a written yes or no before you sign.

Who sets the price? The marketing plan includes “pricing activity,” so S&S does, and the Moriarty Murders print edition is listed at £13.99 on Amazon UK, which is a lot for a print-on-demand cozy mystery.

You also give up the final say on title and cover, and while they’ll work closely with you, their team uses market insight to decide. That’s fine if you trust their team, and less fine if your covers already convert.

Who should take it

I’m not saying nobody should sign. If you can’t afford a good editor and narrator and would like a Big 5 name on the spine, 50% of net with no upfront cost beats the 25% you’d get almost anywhere else in trad, and a debut with no readers loses little by trying.

The mistake is moving a series that already earns, because you’d be swapping 70% for about 35% on books whose readers found them without any help from S&S.

Before you sign anything

• the rights term, plus a reversion clause with a sales threshold, in writing

• the print and audio royalty rates

• whether the ebook goes into KU, and at what price

• what happens to your series if the imprint closes

• whether you can keep the ebook rights and license only audio and print

Or do what Gatland did: give them something new, and keep what already pays you.

A few months ago I published my KDP Dashboard Autopsy Prompt.

The idea was to take my KDP exports, give them to an AI model, and ask it to look at the business without trying to make me feel good about it. Where does the money actually come from? Which books still have momentum? Where does a series lose readers? Am I spending money on books that stopped earning it back months ago?

I still like the prompt. But I ran an updated version on my own data recently and realized there was something fairly important missing.

Not another formula. Context.

KDP can tell me what happened. It is much worse at telling me what I was doing when it happened. That matters more than I thought.

I can look back and see that KENP jumped in a particular month. Fine. Why?

Was I running a promotion? Did I increase Amazon Ads? Did I send a newsletter? Was that when I changed the cover? Did a reviewer or BookTuber mention the book? Was there a sequel launch around the same time? Did I drop the price for a few days?

Maybe I remember if it happened last month. Six months later? Good luck.

And if I don’t remember, the AI doesn’t know either.

It can still give me an answer, of course. That is part of the problem. Give a model a sales spike and enough surrounding data and it can usually come up with a very reasonable explanation for it.

Reasonable does not mean true. I ran into a version of this while testing the new Autopsy Prompt.

One result showed that Kindle Unlimited represented a very large percentage of my recent Amazon royalties for a specific series. On paper, it looked like a strong argument against leaving KDP Select. Then I looked at the actual dollars.

The income for this series had fallen so much that KU could represent most of what was left while still earning me only a few dollars a day. The percentage was correct. The conclusion I initially drew from it wasn’t.

That was easy enough to fix in the prompt. Important percentages now need the actual dollar amount next to them, along with the recent trend. But that sent me down a different path.

What else am I asking the model to work out from incomplete information?

Quite a lot, apparently.

Suppose I changed a cover in March, spent $250 on a promotion in April, switched off Amazon Ads in May because they were going nowhere, and delayed the next book by six weeks in June. All of that could matter when I look at the sales curve later.

None of it is going to be obvious from a KDP export.

Yes, I could probably reconstruct most of it. Amazon Ads has the campaign history. Meta has the spend. MailerLite knows when I sent emails. There are old covers somewhere in Canva. My calendar probably has launch dates.

But am I really going to dig through five different systems six months from now because I want to understand why sales dipped for three weeks in June?

Probably not.

So I think the answer is much simpler. I need to keep a publishing log. Not some giant spreadsheet with 47 columns that turns into another job. Just a short record of anything that might matter later.

Once a week, write down what changed.

How much did I spend on ads and where?

Which book was I pushing?

Did I run a promo?

Change the price?

Send a newsletter?

Launch something?

Delay something?

Change a cover or blurb?

Pause a campaign?

Leave KU?

Add another retailer?

Get featured somewhere?

Have some random KDP problem that seemed annoying but temporary?

Most weeks the entry might be three lines. That’s fine. What matters is that six months later I can look at a weird movement in the data and have something better than, “I think maybe I did a promo around then.”

I would probably keep each entry to five things:

Date. What changed. Which book or series. Money involved. Why I did it.

For anything that is actually an experiment, I would also write down what I expected to happen. Something like:

“Started $10/day Meta campaign to Book 1. Not expecting the campaign itself to be profitable. I want to see whether acquiring more Book 1 readers works once series read-through is included.”

That is the kind of note I wish I had for a lot of things I’ve tried over the last year. Because when I come back later, I don’t just know that I spent money. I know what I was trying to achieve. Then I can compare the result with the original idea instead of inventing the strategy afterwards.

I’d probably do a slightly bigger note at the end of each month as well. Total ad spend by platform. Any launches or delays. Major promotions. Distribution changes. Anything unusual. Nothing fancy. The point is just to save the context while I still remember it.

That’s the thing historical publishing data loses. KDP will happily tell me how many pages were read on June 14. It won’t tell me that I switched an ad campaign off on June 10 because the CPC had gone stupid.

It might show a drop in Book 1 sales, but not that Book 2 was supposed to launch that month and didn’t.

It might show an old title suddenly waking up for a week, but not that somebody with a decent newsletter happened to mention it.

And this gets much more useful once AI enters the picture. If I give a model only the KDP exports, it can find patterns. If I give it the exports plus six months of notes about what I actually changed, it has something much closer to a business history.

It still can’t prove that the new cover caused the sales increase or that the newsletter caused the KENP spike. But at least it knows those things happened. It can compare dates, look for similar patterns elsewhere, and tell me where the evidence is strong and where I’m probably just seeing coincidence. That is much better than asking it to fill in the blanks.

So this is the part I missed with the first Autopsy Prompt.

I spent a lot of time thinking about how to make the AI analyze my publishing data better. I should also have been thinking about what information I need to preserve so there is actually something useful to analyze later.

I’m going to start keeping that log. Five minutes a week should do it.

Because KDP is very good at remembering how many books I sold. It has absolutely no idea what the hell I was doing at the time. And six months later, neither do I.

Subscribe now

I published a three-pass AI editing system on this newsletter a while back. Pass 1 handled scene-level story and prose, Pass 2 handled scene-level copyediting, and a third pass, which I called the Dead Metaphor Pass, read the whole manuscript afterward hunting for repeated imagery and physical tics.

I ran all three on real manuscript, not test snippets, and the first two held up well. Scenes got sharper; the copyedit pass caught things I’d stopped seeing after the fifth read-through, without flattening my voice into something blander along the way.

The Dead Metaphor Pass was the one that changed the most once I started running it on real chapters.

The rule I’d written for it was mechanical: if a gesture, image, or physical tic showed up three or more times across the manuscript, the model flagged it and rewrote every instance past the second. Clean logic on paper, too blunt in practice, because the rule treated frequency itself as the defect. It couldn’t tell me which occurrence was actually weak. Sometimes the first use of an image is the flat one and the fourth is the best line in the chapter. Sometimes a character rubbing her jaw four times in three hundred pages is characterization, not a tic that slipped past me. A rule firing on the count alone can’t see any of that.

So I rewrote the rule: the pass that used to trigger an automatic edit at instance three now treats three or more as a reason to look, nothing more, because frequency alone should never force a change. That’s the one number worth sitting with: the old version rewrote everything past occurrence two by default; the new version rewrites nothing by default, and asks the model to justify each flag on what the repetition is doing, not on how often it happens.

Rewriting that rule surfaced a bigger question. The three-pass system could tell me whether a scene worked and whether the prose was clean, and once revised, whether an image or gesture had gone stale across four hundred pages. What it never told me, not once, was whether the novel worked as one complete story: whether the setups paid off, whether the middle actually escalated instead of repeating itself with new scenery, whether the ending resolved the thing the book had actually been about, or something merely adjacent to it.

No amount of scene-level polish answers that, because it isn’t a scene-level question; the system is now four passes instead of three, with a new manuscript-level structural audit sitting between the copyedit pass and the renamed pattern pass. What each one does, why the reordering matters, and the full prompts for all four are below.

Read more

If ACX won’t take you, the real decision isn’t whether to use an aggregator. It’s which one. Author’s Republic and Voices by INaudio, still called Findaway Voices by most people out of habit, are the two platforms that actually solve the problem for a non-US author, and the choice between them is where things get specific.

Both platforms will get you onto Audible without an ACX account, and both pay eighty percent of net royalties while keeping twenty. Neither charges anything to join, which means the royalty rate won’t help you decide either, since it’s identical on both sides. Both are also rated “Excellent” by the Alliance of Independent Authors, so neither one is a trust question, just a details question.

The real differences show up in the details.

Start with the money itself. Author’s Republic pays out through PayPal once you clear ten dollars in a monthly period, while Voices by INaudio generally sits at a hundred-dollar threshold. For a book that’s selling well, that gap barely registers, but for a backlist title moving a few copies a month, it’s the difference between getting paid every month and watching a balance sit untouched for most of a year.

Then there’s Spotify, where the two platforms split. Author’s Republic’s say that they distribute to Spotify and getting added takes one email to support. Voices by INaudio can reach Spotify too, but only since splitting from Spotify itself in August 2025, and now it takes either their retained retailer relationship or a separate Spotify for Authors account. One extra step isn’t a dealbreaker, but if you want your whole catalog behind a single login, Author’s Republic gets you there in one move and INaudio doesn’t.

AI narration turned out to matter less than I expected going in. Voices by INaudio only accepts it from three named providers, each exporting an unmodified file in one specific format. Google Play Books and ElevenLabs are two of them; Spoken Press, the tool narrating Rising Storm, is the third. I checked that pipeline myself, on both the Spoken side and the INaudio side, and it holds: Spoken exports straight into the format INaudio wants, which makes this a real, working path instead of marketing copy. Author’s Republic skips the named list entirely, with a simpler policy: disclose the AI use per the Audio Publishers Association’s guidelines, and the file goes through regardless of which tool made it. Either way, both doors are open to me.

Contracts show something else, since Author’s Republic writes a seven-year term into its agreement, matching Audible’s own term length, but you can walk away anytime with sixty days’ written notice. Voices by INaudio skips the long-term paperwork and lets you pull a title whenever you want, with about thirty business days for the retailers to catch up. Neither is a trap, just a different bet on what an author is willing to sign.

None of this changes the basic fact: if you’re in Poland, or anywhere else ACX won’t touch, either platform gets your audiobook onto Audible without an account you can’t open. The choice between them comes down to how fast you want to get paid, and whether a seven-year contract you can exit in sixty days bothers you more than no contract at all.

I’m sticking with Author’s Republic for now, mostly for the Spotify convenience and the lower payout floor. Still making audio narration with Spoken. I’ll update you soon on how the experiment goes.

R.L. Davennor published a post last month with a title that does most of the work: she made a low six figures from her books in 2024, and she is now working two day jobs, picking up Instacart shifts and starting an OnlyFans to cover the bills. She names the causes she can see, the TikTok blackout, the algorithm changes, an economy that stopped buying, and she is honest about the part that has no visible cause at all. She kept publishing three books in eleven months, she posted two reels a day, and none of it moved the number.

I read it on a morning when I was going through my own royalty history for other reasons, and the shape was familiar enough to be uncomfortable. April 2025 paid me $5,182, May paid $5,065, and nothing since has come close, with my 2026 monthly average sitting around a thousand.

The instinct is to read those two stories as the same story, where an author does well, the market turns, and the author does worse. That reading is wrong, or at least it is wrong for mine, and the difference matters more than the sympathy does.

Davennor knows what she built. She built an audience on a platform, the platform changed, and the audience stopped arriving. That is a loss with a mechanism attached to it, and a mechanism is a thing you can reason about, argue with, mourn properly and eventually replace.

I did not know what I had built, which is the whole of my problem, and it took eighteen months to see it.

Read more

I hired an AI coworker in May. Three days in, it told me my Facebook ads were reaching an audience I hadn’t chosen.

Not underperforming, not “needs optimization,” wrong audience. I ran Meta’s Advantage+ targeting on my ad account, the version where you hand Meta a seed audience and let its algorithm expand from there, and Viktor pulled the campaign data and found that only 48% of clicks were landing inside the audience I’d actually set up. The other half went wherever Meta decided to send them.

I’d been running that campaign for months, reading the weekly reports, and I never once cross-referenced click destination against my own targeting settings, because why would I. The dashboard shows spend and clicks, and hands you a CTR that looks fine on its own, but none of that tells you that half your budget went somewhere you didn’t send it.

That’s the first thing Viktor did that I couldn’t have done myself in the time I had, and the reason wasn’t that the analysis was hard. I wasn’t going to run it, because I had a business to write books for.

What Viktor actually is, for anyone who hasn’t set one of these up

Skip this section if you already know it. Viktor is an AI agent connected to my actual accounts, not a chatbot I paste numbers into, and that distinction is the whole point. It has API access to my Amazon Ads console, plus Meta, MailerLite, and my KDP dashboard, and it runs on a schedule, a Meta campaign check every morning at 10am, a full attribution report every Friday, alongside just answering when I ask it something in Slack. Setting up the Amazon side alone meant navigating three separate Amazon portals that don’t talk to each other cleanly, since Amazon splits developer access from advertising access, and support access sits behind a third wall entirely. Viktor did that once, and it stayed done.

A chatbot tells you what to check, and you have to remember to ask. An agent with account access checks it on a schedule, whether you remembered or not.

The finding that actually changed how I spend money

Every indie author I know assumes their ad spend is doing something, even when they can’t say exactly what, and I assumed the same thing. So when Viktor reconciled my Meta ad spend against Amazon Attribution data for my best month of the year, I expected to see paid ads driving a meaningful chunk of revenue, even 15%, even 10%.

$5,065 in revenue that month, and $52 of it traceable to Meta ads.

I want to be precise about what that number means, because it’s easy to round it into something dramatic and wrong, and it doesn’t mean advertising is worthless. It means my specific Facebook campaign, on my specific books, in that specific month, produced almost nothing measurable next to organic and Kindle Unlimited discovery, and I’d been treating that ad spend as central to my results for over a year when it turned out to be a rounding error.

I turned the campaign off within the week, and nothing in my sales moved.

The gap I didn’t notice for 44 days

The third thing worth mentioning isn’t a diagnosis, it’s an alert. Viktor’s weekly scorecard pulls ad spend and MailerLite subscriber counts together with Amazon Attribution into a single view, and one Friday it flagged that I’d gone 44 days without sending a newsletter. I run three separate lists under different pen names, and I was sure I’d sent something in that window. I hadn’t.

Nobody catches that by feel, because it isn’t a feeling, it’s arithmetic someone or something has to actually run.

Where I’d push back on the hype

I don’t want to write the version of this post where the AI coworker is magic and I just sit back, because that’s not what happened and it isn’t honest. Every one of these findings still required me to decide what to do with it, kill the campaign, fix the newsletter cadence, question an assumption I’d been carrying for a year. Viktor surfaces the problem, and it doesn’t run the business. Two guardrails I built in from day one, and would tell anyone setting this up to build in too: nothing customer-facing ships without me approving it first, and no number goes into a report without a source I can trace back to the raw export.

Of the roughly 91 hours Viktor’s own usage report says it saved me over 30 days, most of that came from me actively working with it in conversation, not from unattended automation running while I slept. The scheduled jobs matter, and the ad monitor and the weekly scorecard do run without me. But the bulk of the value came from directing an agent that could act, not from a system I could ignore.

If you want to try this yourself

I’m not neutral here, so I’ll say it plainly: the link below is an affiliate link. If you subscribe through it, you get $50 off your first purchase, and I receive a commission.
I wouldn’t put my name on it if the audience-leak finding and the ad reconciliation above hadn’t been real, but you should know the incentive going in.

Try Viktor

Start with whatever account is bleeding money you haven’t looked at closely, since mine was Meta and yours might be somewhere else entirely.

A literary agent I read, Abigail Fenton, published a piece called “No, You Can’t Write Your Novel With AI.” It is well argued, it is honest about being personal opinion instead of industry policy, and I think it is wrong in ways that matter, but the part that stuck with me was the ending, because it gives away more than she meant it to.

She closes with her five-year-old, and the story goes like this: friends went on a family bike ride, ice cream at the end, and the daughter couldn’t come because she has never been willing to put in the time to learn to ride. “If you want to ride a bike, you’d better get in the saddle and learn.”

I have looked at that paragraph for a while now, because every parent I know has solved this exact problem. There is a child seat that bolts over the rear wheel, there is a trailer that hitches to the seatpost, there is a tag-along half-bike that lets the kid pedal without steering or balancing. Four options, roughly forty to three hundred dollars, all of them at your local bike shop, and with any of them the five-year-old gets the ride and the ice cream and, incidentally, five miles of watching how riding a bike actually works.

The analogy chose itself, and it does not say what she wants it to say. It says: when someone can’t do the thing yet, you can either build the thing that carries them, or you can let them stay home and call it a lesson.

The ghostwriter problem

Her strongest argument is legal, and it is the one I take seriously. AI-generated text is not copyrightable, so an author cannot grant a publisher exclusive rights to something they never owned, and publishing contracts contain a warranty that the work is your own creation. If a machine wrote it, you cannot sign that warranty. Fine, and that is an accurate description of where the law sits today.

But she then slides from a contract problem into a definition of authorship, and that is where it falls apart, because publishing already has a very large exception sitting in plain sight. Celebrity memoirs. Politician autobiographies. Business books by executives who have never written a paragraph in their lives. A professional ghostwriter does one hundred percent of the word-picking that Fenton says makes someone an author, and the byline still belongs to the person who supplied the life, the ideas and the voice. Nobody calls this fraud, because it is a category with its own rates and its own quiet etiquette, and it has coexisted with the warranty clause for a century.

So copyright law already accepts that authorship is not the same as typing, and the industry accepts it too, as long as the fingers on the keyboard belong to a human being who has been paid and asked to stay quiet.

Which means the objection to AI is not really about who did the writing, it is about who deserves the status.

Which is it

Here is the asymmetry I cannot get past.

When a language model produces something good, the industry position is that no human authored it, since there is nothing to copyright and nobody to credit, and no art was involved anywhere in the process. When a language model produces something defamatory or plagiarised, the position flips instantly: the human who prompted it is responsible, is liable, should have checked. Publishers ask you to warrant that the text is your own creation and simultaneously insist that it can’t be.

You cannot have both, and the choice is not subtle. Either the person steering the machine is the origin of the work, in which case they own it and answer for it, or they aren’t, in which case go find someone else to sue. Every legal system that has looked at this has quietly chosen responsibility without granting credit, and that isn’t a principle, it’s a convenience.

She is right about AI editors

I want to give her the strongest point she makes, because I ran into it myself and it cost me months.

Ask a model to edit your manuscript and it will edit, which is the whole problem. It has no way to say “this chapter is fine, leave it alone,” so it manufactures suggestions to justify its own existence, and because editorial advice is not verifiable, and you cannot check whether merging two characters would in fact be better, you have no way to tell useful notes from generated ones. Fenton describes it as sounding like an answer while being nonsense, and that matches what I saw.

What she gets wrong is the conclusion, because this is a tooling failure, not a property of the technology, and it is fixable.

What worked for me was refusing to ask the broad question. “Edit this chapter” gets you the same thing as “write me a book,” so I went and read the manuals instead, three of them, and not books about how to write: Susan Bell’s The Artful Edit, Dreyer’s English, and Amy Einsohn’s The Copyeditor’s Handbook. I read them the way an engineer reads API documentation, looking for nameable operations instead of taste. What came out of that is the reason human editors price developmental work and copyediting differently: they are separate layers that ask separate questions, and running them at the same time is why “edit my manuscript” fails.

So I split it into three passes, each one finished before the next begins. Pass one works a single scene at a time on pacing, interiority, dialogue subtext and whether the emotional beat on the page is the one I think is there. Copyediting comes second on that same scene, the four Cs plus an adverb audit and a genre-cliché list I had to build myself for military sci-fi. Only the third pass runs across the whole manuscript, once, and does nothing but hunt repeated images: every sensory detail or physical tic used three or more times to signal the same thing, logged, then replaced from the third instance on. That last one catches what I stopped being able to see, because I had been inside the book far too long and the model had not.

I wrote the whole method up separately, and the three prompts are available to paid subscribers here.

None of this removes me from the process, which is the point Fenton and I are actually arguing about. The passes annotate and flag and propose, and I still decide whether any given replacement is better than what I wrote, and after a few hundred inline notes you start catching your own patterns in the next draft before the model does.

That is a long way from a human developmental editor, and it is also a long way from nothing, which is what most authors can actually afford, and Fenton says so herself: she’d love a more cost-effective way for authors to get editorial input. She started her Substack partly for that reason. I just think she has decided in advance that the cost-effective way cannot be this one.

The people who can’t write

The line that made me put the coffee down: “if you do not have the writing skill to write a book, then perhaps you do not get to write one.”

Storytelling is older than writing by something like thirty thousand years. Every oral tradition that ever existed was built by people with no letters at all, and we only have Homer because somebody eventually wrote down what a non-writer composed. Skill with sentences is one delivery mechanism for narrative, and it happens to be the one that publishing sells, which is why an agent would mistake it for the whole art.

I know people whose imagination I would trade mine for and whose sentences are a mess, some of them dyslexic, some working in a third language, some simply never taught. Under Fenton’s rule they don’t get to have their story exist, and the world is not better for that, it’s just quieter in a way that happens to protect the people who already have the skill.

And the tools argument cuts deeper than she allows, since spellcheck is a shortcut, Grammarly is a shortcut, Scrivener’s structure view is a shortcut, and a copyeditor who silently fixes your comma splices for eighteen months is a very expensive shortcut. Nobody has ever said a dyslexic writer using dictation software didn’t really write their book. The line always seems to fall just after whatever the person making the argument already uses.

The random number

Her best rhetorical moment is a quote from Claude’s own documentation, where the model explains that when either “overcast” or “grey” would do, the choice comes down to a random number. She wants authors who care intensely which word it is.

So do I, and that is the entire argument for keeping a human in this, and she’s presented it as an argument against.

A model with nobody watching produces exactly the slop she is describing, and there are hundreds of thousands of those books on Amazon right now, all averaging toward the same texture, so the difference between that and something worth reading is a person going through it deciding that no, it’s not grey, it’s the colour of the sky over a firebase at 0500, and that this paragraph is doing nothing and comes out. Word by word, decision by decision, the same job she’s describing, just starting from a draft instead of a blank page.

You can tell the difference in the reading, which is why I’m not worried about the flood in the long run, since readers do not owe anyone patience.

What I’m actually asking for

I don’t want publishers to accept unedited machine output, and I don’t think they will or should. What I want, and it feels like a small ask, is for the argument to stop pretending it is about the writing when it is about gatekeeping, because those need different answers.

If a publisher’s policy says do not upload any part of your manuscript to any LLM – and Fenton reports that some do, which she flags herself as extraordinarily broad – then that policy is not protecting the quality of anyone’s sentences. It is protecting a screening function, and the screening function is worth understanding. Traditional publishing’s service to readers was deciding what was good enough to print, and its service to authors was access to shelf space, the second of which self-publishing dissolved fifteen years ago. Tools that let unskilled storytellers produce readable manuscripts threaten the first, and the response has been to add a new admission requirement at precisely the moment the old one stopped working.

Maybe that’s the wrong read and there is a version of the AI policy that really is about the writing. If someone can show me a publishing house that will take a manuscript from a dyslexic storyteller who dictated and machine-drafted and then spent two years rewriting every line by hand, I will drop the whole argument. I have not found one, though I admit I have been looking from the outside, and I don’t think the question has been put to anyone seriously.

The daughter still can’t ride a bike, and there’s a trailer in the garage.

My author websites were named after books, which sounds harmless, and meant that eighteen months of publishing built nothing I could carry forward.

One was thelightningwar.pl, the other silentrepublic.pl, and both were built the way a writer builds things. I finished a series, wanted a home for it, and registered a domain with the series name on it. Then I did it again. Every new series would have needed a third site, a fourth, a fifth, each one starting from zero authority and zero traffic.

Readers don’t work that way. A reader who finishes book three of The Lightning War and wants more doesn’t search for a series title she half-remembers, she searches for the author name, and until last week, if she searched mine, she found nothing I owned.

So I spent an evening rebuilding from scratch, and both sites, milesphoenixbooks.com and alexradec.com, are live now. Before writing a line of HTML I ran the research properly: a survey of what military SF and post-apocalyptic author sites actually do, plus conversion and technical benchmarks for indie sites in 2026. Some of it confirmed what I assumed, a larger portion contradicted me, and one finding cost me a design I’d been fond of for two years.

The number that made this urgent

In May 2025 my books earned $5,065, and Amazon Attribution says my Meta ads drove $52 of that.

I’d been running paid ads as though they were the engine, when they were a rounding error. Whatever sold those books was happening somewhere I couldn’t see and couldn’t control, which is a polite way of saying it happened inside Amazon’s algorithm. When intake collapsed, and my revenue fell from $5,182 in April 2025 to $565 in July 2026, down 84%, I had no channel of my own to compensate with: no traffic, and no way to reach past readers except two mailing lists sitting on 408 and 390 subscribers, both of which I’d treated as an afterthought.

The Written Word Media 2025 survey of 1,346 indie authors found that the cleanest predictor of monthly income isn’t catalog size or ad spend, but email list size. Authors earning above $10,000 a month average over 18,000 subscribers. I have under 800 across two pen names, eighteen months after my first novel went up

That, and not aesthetics, is the whole argument for rebuilding. My old sites looked fine, they just weren’t doing the one job a website does that Amazon will never do for me, which is to turn a stranger into someone whose email address I own.

If you take nothing else from this, take the ordering: your website’s job is not to sell books, because Amazon closes better than you will. Its job is to capture the reader before Amazon gets her, so that when your rank drops, or a category shifts, or a new book lands in a dead week, you still have a way to speak to the people who already like your work.

Read more

Author’s Haven is my beta reader platform, and the pitch is simple: authors upload a manuscript, readers answer survey questions chapter by chapter, and the author ends up with structured feedback instead of “loved it, keep going.” There’s a free tier that stays free, not a trial that expires on you, and I built the thing because I needed it for my own books.

In one stretch of a few weeks I got two sets of messages that, read side by side, sound like they’re describing different products.

The first came from an author on the free plan, subject line “Looks Like A Con.” I’ll paraphrase lightly, because I don’t want to put words in anyone’s mouth. There are, he wrote, no readers on the site, so the whole thing is a waste of time. He has 56 chapters and has to upload them one at a time, when he could just email an attachment. The software is primitive, and questions had to be answered three or four times because his input kept vanishing. Then the line that actually stuck: I read your article about how you studied all the other beta reader sites to build the best one, and what a disappointment.

Then a support ticket arrived from the same address, subject in capitals, demanding that I delete the account and all associated data immediately and confirm by email. The body called me a cheat and a beggar with nothing to offer except begging for money, and the software crap, and the time spent an inordinate waste.

I deleted the account within the hour, because that part isn’t complicated, whatever else I think of the message.

The second person is a beta reader who writes as well. Her first message opened with how much she was enjoying the platform, and then singled out inline comment categorisation as something she hadn’t seen on any of the other sites she’d used. The bug report came after that: whenever she switched to another window, the page refreshed and everything she had typed into the survey was gone, consistently, in Firefox. She was tabbing away to check the forum or to look at the story map, and coming back to an empty box. Her question was polite to the point of generosity – is this intentional, or could it be a bug?

It was a bug, and the page had been refetching in the background when it lost focus, and Firefox in particular blanked the open survey while it did so. Two changes went out the same day: survey answers now save to the browser as you type, and the background refetch stops while a survey is open. She wrote back to say the whole beta reader group had tested the fix on mobile and on desktop, that several of them had hit the same thing, and thanks.

Here’s the part I keep turning over. *Both of them hit the same bug*, lost survey input, almost certainly a variant of the one thing, and one report ended with the platform being a con run by a beggar while the other ended with a fix shipped in a few hours and a group of readers who trusted the tool more that evening than they had that morning.

Same bug, two different products

Both of them were right about the facts. The software was losing people’s typing. What separated the two messages was never accuracy, it was what each of them had come for in the first place.

The difference had a price, and it wasn’t paid by the polite one. She told me the browser, the trigger, and what she was doing when it happened, so I could reproduce her version in minutes and ship the fix. He told me his input kept vanishing and then closed the account before I could ask a single question, so his variant of the bug may still be sitting in the code, waiting for the next author who writes long survey answers in whatever browser he was using. The furious report was worth less to me than the friendly one, and it cost more to receive.

She came to read manuscripts and be useful to authors, so the platform’s job, from where she stood, was to stay out of her way. When it failed at that, the gap between what she wanted and what she got was one broken behaviour, and describing it precisely served her own interest, because she wanted to keep using the thing.

He came for readers. He had 56 chapters and no audience, and what he needed from me was traffic, so everything else was noise. The upload flow annoyed him and the survey bug annoyed him, but neither was the actual complaint, which was that a free account could not hand him a waiting queue of beta readers for a 56-chapter book, and against that expectation no interface work I could have shipped would have registered. The bug wasn’t his problem with the product, it was evidence for a conclusion he had already reached.

Which is why “you can’t make everyone happy”, true as it is, is the lazy version of this lesson. The useful version is that identical feedback means two different things depending on what the person walked in wanting, so my job is to work out which want I’m looking at before I decide what the message is worth.

What I do with it in practice

On the 56 chapters he was wrong about the facts and right about the product. Bulk import already exists, and has for a while. You open your book, press Import, drop in an EPUB or a DOCX, and it splits into chapters on its own, reorderable afterwards, and pasting a whole manuscript works too. There’s also Send to Kindle, if you’d rather the reader got the book as one file, with feedback still coming back chapter by chapter. He never found any of it, and he had 56 chapters in front of him and a page that looked like it wanted them one at a time, so he did the arithmetic and concluded the software was primitive. A feature nobody finds might as well not exist. That’s a discoverability problem, not a primitive one, and it’s worth knowing which of the two you’re being shouted at about.

Being insulted and being wrong are separate conditions, and I read my inbox with them kept apart.

On the reader supply problem he was right, and that one is much harder, because a beta reader platform is a two-sided market and the empty side is readers. That’s the reason I’ve spent recent weeks on the reader experience instead of more author tools: categorised inline comments, autosave, a “load older feedback” button that someone wrote in specifically to thank me for. There’s a public shelf at authorshaven.io/shelf where a cover and the first 500 words of chapter one go live for anyone browsing, opt-in from your book settings, and most authors here bring their own beta readers and use the site for the feedback part. Authors turn up on their own, since they need something badly enough to go looking. Readers turn up only if the work is pleasant, and they leave the moment it stops being pleasant. So every hour I spend on their side of the platform is an hour spent on the thing authors are actually asking me for, even when it doesn’t look like it from the author’s side of the screen.

The “beggar” line still irritates me, so let me be plain about it. The free tier on Author’s Haven is permanent, and it isn’t a countdown to a card form, whatever he assumed. There’s a paid tier on top of it for authors who want more pen names and the AI features. I want manuscripts and readers inside the system, which is what the free tier is for, and servers and my hours cost money, which is what the paid one is for. If charging for software is begging, then every author selling a novel on Amazon is running the same racket.

Where the two of them were standing

The line that landed hardest wasn’t the insult, it was the disappointment. He had read my article about studying every beta reader platform I could find before building mine, and he measured what he found against the platform that article implied.

She had read nothing about it. She showed up, started reading somebody’s manuscript, and judged the thing by whether it let her do that.

So one of them arrived with a picture in his head and compared the product to the picture, and the other arrived with a task and compared the product to the task. That difference decided everything that followed, including how they each described a bug that was, underneath, the same bug. It isn’t a story about a good user and a bad one. It’s a story about how much of an opinion is already formed before anybody logs in.

Two people, one week, one bug, and I fixed it for one of them and deleted the account of the other, and both were the correct response to what was in front of me.

If you write, or if you read for people who do, Author’s Haven is at authorshaven.io, and the free tier stays free. Tell me what’s broken, in detail if you can manage it. I would much sooner hear it in a bug report than in a subject line.

After the Pangram piece went out, a reader left a long comment, and then muted me. I’ve read it four times now, and it is the best argument against my position that anyone has made, which is awkward, because it also happens to prove the thing I was worried about.

Her case, compressed: the who matters. You look at the diplomas on a doctor’s wall before you take his advice, and a Rothko sells for eighty-five million because of whose hand held the brush, not because coloured rectangles are hard to make. Readers want to know who is behind the words, and if a writer says he uses AI in his workflow, some of them will pass. Detectors, she wrote, are all that readers have, and running someone’s sample chapter through one is a small piece of power in a market that took a lot of power away from them this year.

I agree with one half of that. Readers are owed the information, and anyone who wants to skip me over it should skip me, since being skipped knowingly beats having someone feel tricked three books later. The other half, that the who is what makes the work worth reading, is where I think she is wrong, and I’ll come back to it.

Here, though, is the part I can’t get past. She never ran my article through a detector, because she didn’t need to. I told her, in print, unprompted.

And the detector she is defending is the one from the article she was replying to, the same tool that stamped 100% on prose nobody could flag by reading it. Her empowerment, then, runs on an instrument she has no way to check. There is no appeal, no error bar, and no account, however thin, of what the machine actually saw. She trusts the score because the alternative is trusting writers, and writers have given her plenty of reason not to.

So look at what actually happened, in order. I disclosed, she read the disclosure, she left, and nothing about the sentences on the page changed between the moment she liked the work and the moment she didn’t.

That is the incentive structure, and it’s brutal in a boring, procedural way. Stay quiet and you are clean until an algorithm says otherwise, and if it ever does, you get to be loudly outraged about false positives, with the numbers on your side. Say something and you have handed over the evidence yourself, permanently, with no false-positive defence available to you. The only voluntary signal in the whole system, a writer telling you what he did, is the one that gets punished hardest.

Run that forward a year and ask which writers are still disclosing. Not the ones who watched what it cost, I think. The norm eats itself, and what is left is a market where everyone is silent and everyone is scanning, and the scanners are wrong at a rate that nobody publishes.

I keep coming back to the diplomas, because I don’t think that is how any of us actually choose a doctor. Standing in a waiting room reading the frames on the wall is not something I have ever done. I ask around, find out who my friends were happy with, then go to the one people say is good, and no doctor has ever been crossed off my list for training in Karachi instead of Boston. The wall of paper is decoration; the reputation is the real instrument, and it is built out of hundreds of people reporting what happened to them afterwards.

Books work the same way, only more so. Nobody has ever bought a novel because of the author’s credentials, and if the diploma test applied to fiction, most of the shelf would be empty.

The Rothko is a worse example than she thinks, because she chose the one market where price and merit came apart decades ago. Nobody pays eighty-five million for the pleasure of looking at coloured rectangles. They pay for a name, a provenance chain, a tax position, and a thing to hang where visitors will see it, and the proof is that an identical canvas by an unknown hand is worth a few hundred euros and the same canvas relabelled as a Rothko is worth a fortune. There are painters alive right now, better than most of what hangs in the auction houses, selling nothing, because the market that trades on the who is not looking at the work at all. That is the argument for judging the object in front of you, not the argument against it.

Which is the same trade she is making with my article, in miniature. Her filter is not “is this any good,” it’s “was a machine anywhere near it,” and those two questions come apart the moment you apply them, since she would take a worse post written by hand over a better one written with help and count that as a win. She is buying the label. So is the collector.

The honest version of her position isn’t “detectors give readers power.” It’s closer to this: I don’t want to read AI-assisted work, I have no reliable way to identify it, so I will act on the confessions I get and take everything else on faith. That’s coherent, and I understand it, but it produces a world where the writers who tell the truth are the only ones anybody ever filters.

This is where she and I part company for real, and I’ll say it plainly instead of faking a doubt I don’t have. If I read a book and it moves me, I do not care how it was made. No one has yet explained to me the mechanism by which a paragraph I loved on Tuesday becomes worse on Wednesday when I learn which tools were open on the writer’s desk. The words did not change. What changed is a story I am now telling myself about the person, and I would rather have the book.

What I know is that I am now slightly worse off for having written the sentence “I use AI in my workflow,” that I will write it again anyway, and that I cannot recommend anyone else do the same with a straight face. Show me a version of this where disclosure buys a writer something, anything, and I will switch sides that afternoon. Nobody has yet.

Every few months an author YouTuber makes the same video, and every few months I watch it, because the question underneath it is a good one even when the answer is bad. The latest comes from Kelsey Humphreys, a romcom author with thirteen novels across three pen names, and she is asking why badly written books outsell well written ones. She is honest in a way most of the genre isn’t, since her last release flopped and she says so on camera, and she spends a minute explaining that the revenue numbers other authors post online are not profit numbers. That minute is worth the twenty-five.

The rest is a list I could have written from memory: publish more, publish faster, run ads, build a brand, get rabid fans, build an email list. Most of it survives contact with the data. But the premise is wrong, and the premise is doing all the emotional work, which is the reason the video exists at all.

Nobody buys a bad book, because buying happens before reading. What a reader buys is a cover, a title, a category placement, a blurb, a price, and some social proof, and none of those six things contains a sentence of your prose. The purchase closes before your writing gets a vote. So the comparison that makes authors miserable at two in the morning, my good book against her bad one, sets two objects side by side that were never judged on the axis the question assumes.

The research is unkind about this. A study of 98 UK bestsellers from 1998 to 2012 checked Goodreads data against copies actually sold, and found that the number of ratings predicted sales far better than the level of those ratings did. Rating count measures how many people read the thing, while rating level measures whether they liked it, and only the first one moves money. A 2021 paper in the American Economic Review goes further, estimating that crowd star ratings shift consumer welfare in book publishing by more than ten times what professional critics manage. Ratings and reviews are visible before purchase, while quality is not, and that asymmetry is the entire mechanism. It isn’t a conspiracy against literary merit, whatever your writing group tells you at eleven on a Thursday.

If prose doesn’t sell the first book, what does the successful indie own that you don’t?

One number under six pieces of advice

Take Humphreys’ list seriously for a moment, because each item on it is correct and each item is a proxy for something she never names.

Quantity holds up. Written Word Media’s 2025 survey, with 1,346 respondents, shows the income curve stepping up around ten published titles and stepping again past twenty-five, while roughly 44% of the sample earns under $100 a month and about 13% clears $5,000. Ten books is not a hobby, it’s a decade of weekends, or three years of doing very little else.

Cadence half holds. The 30-day cliff is real, and publishing every ninety days does keep you inside the new-release visibility window. But the A10 rollout in the summer of 2025 changed what the machine rewards, and ALLi’s Orna Ross, writing in The Bookseller, put it plainly: the weight of evidence suggests the store now prioritises sustained reader engagement over short-term launch spikes. Rachel McLean, who won the Kindle Storyteller Award in 2021, reported no drop in sales from the change, only more rank volatility than usual. A launch spike used to be a strategy by itself, and now it’s an input.

Ads hold up as a mechanism and collapse as a number. Funnelling every advertising dollar at book one and letting the series recoup it is correct, and it is what serious operators do. Humphreys says the big names spend hundreds of thousands, and a handful of them do, but the documented indie reality sits closer to $1,300 a month at roughly 50% ACOS, with a $45,000 month quoted in the trade press as an industry high. That gap matters, since “spend hundreds of thousands” is a reason to give up, while “spend $1,300 badly for four months while you learn” is a decision you can take on a Tuesday.

Brand consistency holds up, and here there is a real paper behind it. Management Science ran Goodreads data from 2007 to 2017 and named the effect categorical stickiness: readers keep filing you under the category your earlier books established, and your most experienced readers are the ones least likely to notice you’ve moved. Humphreys lived this, since she launched a romantasy trilogy under a second pen name while marketing it to her romcom audience from the same Instagram account, and it died. She is now repositioning it as romcom fantasy. The pen name was never the problem, the audience she pointed at it was.

The email list holds up, and it holds up hardest of all. In the same survey, authors with a working list earned around $300 a month against roughly $15 for authors without one. That is not a marginal effect, it’s a twentyfold spread, and it is the starkest correlation in the dataset. Lulu’s 2025 report on 2,065 creators lands in the same place, with every author making $40,000 a year or more sitting on at least a thousand subscribers.

Six pieces of advice, and every one of them is a different way of paying for the same thing. More books means more chances that a reader who liked one will find another, faster cadence means less time for that reader to forget you, and ads at book one only make sense if books two through six earn the spend back. Brand consistency means the next book satisfies the appetite the last one created, and the list means you can tell those readers the next one exists.

The thing being bought, in every case, is read-through.

What I can measure on my own shelf

I can’t run a controlled experiment on the publishing market, but I can read my own dashboard, and book one to book two on my main series converts at 42%. Just under six readers in ten finish the first book and don’t continue, or don’t finish it at all, and KDP won’t let me separate those two failures cleanly, which is a small scandal in its own right.

That single number is the ceiling on everything else. At 42%, a reader acquired for $3 has to earn back that $3 from a book-one royalty plus a 0.42 chance of a book two, then a 0.42 × whatever chance of a book three, and so on down a decaying chain. Push the rate to 60% and the chain roughly doubles in value without a single additional word written or ad dollar spent. Nothing else on Humphreys’ list has that much return per hour of work.

She stumbles into this herself and doesn’t notice. The best story in her video concerns a drop-off after book two in her series, and how she rewrote the ending and the back matter of that book to point harder at book three, and how she later saw a reader post about starting the third one. That is the whole argument, and she tells it as a warm anecdote with a sample size of one, sandwiched between two plugs for her paid blueprint.

Kindle Unlimited makes the stakes plainer, since there the metric is priced in public. December’s global fund was $64.9 million including the All Stars bonus, and the 2025 rate averaged $0.00445 per page, with October hitting a five-year high of $0.005. On a 300-page book that’s about $1.34 for a complete read, so a borrower who stops at 40% pays you 54 cents. The subscription economy meters your read-through every month in full view, and the number authors still argue about online is rank.

The part I’m less sure about

The market data cuts against the simple version of my story. Print was flat in the first half of 2026, down 0.3%, but adult fiction rose 5.7% while YA fiction, the BookTok category that supposedly built all of this, fell 25.6%. Manga was up 45% in the first quarter. Attention didn’t disappear, it relocated, into categories that are almost entirely series-shaped and binge-shaped, which either confirms everything above or means I am reading a manga boom as a read-through story because that is the story I brought with me.

Humphreys tells her viewers to stay a little delusional, and she is right, since it’s the only defensible reason to do this work at all. Be delusional about your prospects, though, and not about your numbers. Prose is the part you control completely, and the part the market checks last.

I wrote “The Indie Author’s Guide to Newsletter Swaps” in February, and I meant every word of it. Newsletter swaps were the cheapest way I’d found to break the Cold Start Paradox, that loop where you can’t get into swaps because you don’t have a list, and you can’t build a list because you aren’t doing swaps.

Then I watched my lists for six months, and what I saw made me rethink the advice.

The numbers on my Miles Phoenix newsletter showed one pattern: open rate of 43%, click rate of 15%. When I send a launch email for the Warfront Dispatch, people buy the book. That list runs on readers who found me through my novels, signed up because they wanted the next one, opened the follow-ups, and actually read what I send them.

My M.P. Blackthorne list showed something different: open rate of 25%, click rate of 4%. Blackthorne was the pen name where I leaned hardest into group promotions and BookFunnel swaps, and while the subscriber count grew faster, the list looked deader by every measure that matters.

Same author, same email platform, same approach to writing newsletters, same MailerLite account, same everything. The only variable was how those subscribers arrived.

What I Mean by “Freebie Seekers”

This isn’t a character judgment, since freebie seekers are simply people who responded rationally to the incentive they were handed. You put a free book on a multi-author landing page with 40 other free books, and they grabbed yours, plus eleven others, because why wouldn’t they? They gave you an email address and you gave them a download link, and whatever relationship existed ended at that moment.

The problem is that this transaction has almost nothing to do with wanting to read your books specifically. The reader didn’t discover you through your writing or your world, they discovered a page full of free stuff, and you happened to be standing on it.

Dan Padavona ran the numbers on this over six months and published them, which took some nerve, because the results don’t flatter anyone. He was adding 200–250 subscribers per month through BookFunnel list builders, which sounds like real, meaningful growth until you look at what happened next. About half never opened a single email, so he deleted those, and of the 589 who remained, only 123 ever clicked a buy link, which works out to 20.9%.

His raw cost per subscriber was $0.15. His cost per subscriber who actually clicked a purchase link: $0.73. And clicking a buy link isn’t the same as buying, so his real cost per paying reader was higher still. It gets worse when you factor in the damage those dead subscribers do to everything else on your list.

The Costs You Don’t See

MailerLite charges by total subscriber count. Not by how many people open your emails, not by engagement, not by clicks, not by a single measure of whether those addresses are worth anything to you. Every subscriber counts the same whether they buy every book you release or whether they’ve never opened a single message you sent. A list bloated with non-openers pushes you into a higher pricing tier for the privilege of emailing people who ignore you, which is a frustrating way to spend money.

That’s the minor cost, and it’s nowhere near the worst of it.

The major cost, the one that compounds quietly, is deliverability. Google and Yahoo monitor how recipients interact with your emails, and they weigh opens against complaints and against how long your messages sit unread. When a large percentage of your list ignores you, your sending domain gets classified as low-value. Your emails get quietly routed to Promotions tabs and spam folders, and the people who suffer most aren’t the freebie seekers who were already ignoring you, but the engaged readers who actually want to hear from you. The engaged 43% starts dropping because your domain got flagged.

ScribeCount published conversion benchmarks earlier this year, and the ordering matters more than any single figure in it. Your own existing subscribers sit at the top, converting somewhere between 5% and 15%. Back-of-book links inside your ebooks come next at 3–8%. Newsletter swaps and cross-promotions manage 1–4%, which puts them barely above paid ads against cold audiences at 0.5–2%, and only one step further along than cold social media, which limps in at 0.3–1%.

Swaps land closer to cold advertising, in other words, than to any channel you actually own. The range is wide because genre match matters enormously. A swap with an author whose readers love the same sub-genre can convert at 4%, but a swap with someone in a loosely related genre drops below 1%, and most group promos land on the wrong end of that range because the matching is approximate at best.

Why the Incentive Structure Creates This

Group giveaways are designed to generate volume, and on that narrow measure they deliver. You join a promo with 40 authors, everyone emails their list, thousands of readers hit the landing page, and you walk away with a few hundred new email addresses. That part of the pitch is true.

But the reader on that page, scrolling past forty covers, is facing a buffet. They’re picking up everything that looks interesting, working through a dozen authors in a single sitting without registering a single name. Some of those readers will end up loving your book, but most won’t remember your name by the time they check their inbox the next morning. You’ll send them a welcome email, and it’ll sit between the welcome emails from nine other authors they signed up for on the same page, unread and eventually forgotten.

One-to-one swaps are better, and meaningfully so. When another author personally recommends your book to their list, that recommendation carries weight because the reader trusts the recommending author, and the genre match is usually tighter because you picked each other deliberately. But even here, the subscriber signed up for a free thing, and they didn’t find you through your work, they found you through someone else’s endorsement of a freebie.

Christopher Schmidt, who’s been documenting his BookFunnel and StoryOrigin experience publicly for years, identified two species of dysfunction in how swaps actually work. “List leeches” join swaps to get their book promoted but never email their own list with their partner’s book. “List stuffers” join 30 or more promos simultaneously, diluting every recommendation to the point of meaninglessness. He found a PNR author with 16,000 subscribers who had 65 active shares in a single week, and most of those shares produced single-digit clicks. The subscriber count was, in every practical sense, a costume, and her actual reach was a fraction of the number on her MailerLite dashboard.

What Builds an Actual List

Read more

Every indie author forum I’m in right now has the same thread at the top. Some version of “Is KU dead?” or “Should I go wide?” or the one I keep screenshotting to laugh at later: “TikTok Shop changed everything and you’re leaving money on the table if you don’t see it.”

The prompt is real. Multiple six- and seven-figure romance authors have publicly pulled their catalogs out of Kindle Unlimited this month. BookTok Times ran a piece in early July citing $230 million in year-over-year royalty outflow from KDP Select romance titles between Q1 2025 and Q1 2026. Maren Cole, an indie dark romance author with 2.3 million BookTok followers, said she went from $40,000/month in KU page reads to clearing $310,000 in a single month after going wide and seeding TikTok Shop affiliates. The Alliance of Independent Authors now says the breakeven point has dropped to around 80,000–100,000 monthly page reads for authors with a meaningful TikTok presence. Two years ago, the consensus was 300,000.

So this is where I’m supposed to panic, announce my new Shopify store, and start courting BookTok creators to push my military sci-fi series.

No. I’m staying in KU. And I think most authors outside of romance should too.

The romance math is real. For romance.

I’ll get into my own genre in a second, but I want to say this first: the romance authors leaving KU are not making a mistake. For their genre, at their scale, with their audience infrastructure, the math works.

TikTok Shop collapsed the discovery problem that used to make wide distribution a grind. A creator posts a “this book wrecked me” video, the algorithm surfaces it to 400,000 people by hour six, the purchase happens inside the app. The author selling wide keeps margin on a $14.99–$19.99 paperback without surrendering exclusivity to Amazon. A signed edition through Shopify at $24.99–$34.99 nets $16–$22 per unit after shipping.

Compare that to the KU math. The current KENP rate sits at roughly $0.0043 per page read, down 11% since 2023. A 250-page romance novel earning 500,000 page reads per month generates about $2,150. The same book sold wide at $4.99 across Apple Books and Kobo returns $3.49 per unit at 70% royalty. Sell 615 copies and you’ve matched your KU income. Add TikTok Shop paperback margin and the comparison isn’t close.

Romance has the audience infrastructure to make this work. BookTok is a romance and romantasy discovery engine. Dark romance, enemies-to-lovers, smut-forward fantasy, bully romance, age-gap. These are the categories where creators have built massive affiliate networks. The supply chain (IngramSpark fulfillment, Shopify backend, creator affiliate commissions at 8–12%) has matured into an actual operational playbook. Firms like Author Amplified charge $3,500–$8,000 to set you up. The path is well-worn.

That path does not exist for military science fiction.

My genre doesn’t have a TikTok Shop flywheel

The audience for my Lightning War saga, my mil-sci-fi series under the Miles Phoenix name: KU subscribers. Not people who stumbled into KU to read one book. People who pay $11.99/month because they chew through two or four sci-fi series a month and KU is how they find the next one. The discovery mechanism is Amazon’s recommendation algorithm. Not BookTok. Not Instagram reels.

Mil-sci-fi readers don’t buy sprayed-edge special editions. They read on Kindle, they read in KU, and the format and the platform are the same thing for this audience.

The Lightning War saga runs five books plus two interstitial novellas and a reader magnet. Average KENP page count for a full-length entry: roughly 450 pages. If a reader starts Book 1 in KU and reads through, that’s 2,250+ KENP pages across the core saga alone. At $0.0043/page, one full readthrough is worth about $9.68 in page-read royalties.

My readthrough from Book 1 to Book 2 sits at 42%. Mid-pack for series fiction. But every percentage point I claw back is pure margin. A reader who enters at Book 1 and reads through to Book 5 is worth almost ten dollars in KU revenue. Getting that reader into the funnel is what KU’s recommendation engine does better than any other platform for this genre.

I had a conversation about this with another mil-sci-fi author last month at a virtual meetup. He’d gone wide six months earlier, moved his whole catalog to Draft2Digital. His exact words: “I thought Apple Books would be my second Amazon. It’s my second Smashwords.” He was considering going back to KU Select. His wide sales across four platforms combined were less than half his old KU page-read income. And this guy has twelve books out.

If I pulled my catalog out of KU and went wide tomorrow, I’d lose the KU recommendation algorithm. That’s the single biggest cost. Amazon’s “also bought” and “readers who enjoyed” carousels are the primary way new readers discover mid-list mil-sci-fi. There is no equivalent on Kobo or Apple Books. Without KU’s algorithmic boost, I’d need to double or triple the ad budget to maintain the same reader inflow. Meta Ads at $5/day and Amazon Ads are already part of my spend.

I’d gain access to Apple Books, Barnes & Noble, Kobo, Google Play. Fine platforms. But mil-sci-fi readership on these storefronts is a fraction of Amazon’s. The data from Draft2Digital and Smashwords shows that science fiction (military and hard sci-fi in particular) is one of the most Amazon-concentrated genres in indie publishing. Going wide for a romance author means accessing a larger, distributed audience. Going wide for me means splitting a small audience across five platforms instead of dominating one.

And TikTok Shop offers me nothing. There is no mil-sci-fi BookTok creator network. No affiliate network. No “this book destroyed me” video format that works for political space opera. I’d be going wide without the demand engine that makes wide work.

The KENP rate compression is real, but context matters

One argument I keep seeing: the KENP rate has dropped 11% since 2023, KDP Select now carries 4.2 million enrolled titles (up from 2.9 million), and individual author share of the global fund compresses as enrollment grows. All true.

But the KU global fund hit $890 million in 2025 and is tracking past $1 billion for 2026. Per-page rates are down, but the total pie is the biggest it’s ever been. For an author with a deep catalog and decent readthrough, the math still works because you’re earning across a multi-book funnel. The KENP rate on any single book doesn’t tell you much. The cumulative page reads across a five-book series with novellas filling the gaps? That’s the number.

Amazon also bumped the KU page-read rate by 4% in April 2026 for romance and romantasy titles with completion rates above 65%. That’s Amazon’s panic response to the romance exodus. Targeted. Probably temporary. But it tells you something about how Amazon views the threat: genre-specific. They’re trying to retain romance. They’re not losing sleep over mil-sci-fi authors going wide because mil-sci-fi authors aren’t going anywhere.

When going wide actually makes sense

Going wide makes financial sense when a few things line up at the same time.

Your genre needs active social commerce discovery. Romance, romantasy, dark fantasy, contemporary YA. These have creators with 100K+ followers who push books through affiliate links. If your genre doesn’t have that ecosystem, you’re going wide into a vacuum.

You also need direct-sales infrastructure. Shopify store, IngramSpark setup, fulfillment partner, an email list big enough to drive launch-week sales without Amazon’s algorithm doing the work. Romance authors absorb that operational complexity because the margins justify it. A mil-sci-fi author selling 700 units across four distribution channels? The operational overhead eats the margin gains.

Release velocity matters too. The romance authors making wide work are releasing six to twelve books per year. They can lose the KU algorithmic boost because they’re feeding fresh titles into TikTok Shop and wide platforms every month. If you publish one or two books a year, typical for longer-form genres, you need each release to work harder for longer. KU’s recommendation engine does that. A wide launch with no TikTok Shop presence gives you a two-week sales window and then silence.

And your per-unit margin on physical sales needs to exceed your KU catalog earnings. Run your own numbers. Take your average monthly KENP reads across your full catalog, multiply by $0.0043, and compare it to what you’d earn selling the same readership as $4.99 ebooks at 70% across wide platforms. For most mil-sci-fi authors I’ve talked to, KU catalog earnings win because readthrough is where the money lives.

If all of that is true for you, go wide. The romance authors meeting those criteria are right to leave KU.

But if you’re writing in a genre where BookTok doesn’t exist as a discovery channel, where your readers subscribe to KU as a lifestyle, where your catalog depth is your competitive advantage, then the “KU is dead” narrative isn’t about you. It’s about a genre-specific disruption being reported as a universal one.

The line I’m watching

I’m not ideological about this. If the KENP rate drops below $0.0035, the per-page economics start to break for long-form series fiction, and the algorithmic benefit would need to be enormous to compensate. We’re not there yet, but I have a spreadsheet with a red cell that lights up if we get close. A sci-fi-focused social commerce channel with real affiliate infrastructure could also change the calculation. BookTok is young. It’s possible that mil-sci-fi or LitRPG content finds its own platform in the next couple of years. My readthrough from Book 1 to Book 2 is the other number I watch. If it drops below 30%, the catalog-depth advantage of KU weakens and selling Book 1 wide at full price starts to look better than feeding a funnel most readers abandon.

None of that has happened. So I’m staying.

The romance exodus from KU is a genre story being told as a platform story. The economics are real, for romance. The TikTok Shop infrastructure, the affiliate networks, the Shopify fulfillment. All real, all built for romance and romantasy. None of it transfers to mil-sci-fi or most other genre fiction without the social-commerce discovery engine that powers the whole machine.

Know your genre. Run your own numbers. And when someone tells you “the math has changed,” ask them: changed for whom?

I scrolled past this announcement. I’m on Substack, so MailerLite’s pricing changes are someone else’s problem.

Then I looked at what “3 automations” means for a typical indie author email setup, the kind with a reader magnet, a welcome sequence, and a new-release notification. And I stopped scrolling.

If you’re on MailerLite’s free plan, you have three weeks to decide what to do about it. What changed, why the automation cap matters more than the subscriber cut, and what your real options look like before the deadline hits.

Three cuts in two years

MailerLite has sliced its free plan three times since September 2025, and each cut came paired with a small feature unlock, as if the unlock made the cut easier to swallow.

Before September 2025 you got 1,000 subscribers, 12,000 monthly sends, and unlimited automations. That was the plan every YouTube channel told you to sign up for, the one that made MailerLite the default recommendation in every Reddit guide and blog post for half a decade.

September 2025 brought the first cut: subscribers dropped to 500, but sends stayed the same and automations stayed unlimited, so most authors barely noticed.

On July 1, 2026, subscribers dropped again to 250 and monthly sends were cut to 2,500, though automations remained unlimited.

August 13, 2026, which is three weeks from today: subscribers stay at 250, sends stay at 2,500, and automations are capped at 3 with a maximum of 5 steps per automation, plus 3 signup forms, one landing page, one website, one digital product.

The subscriber cut to 250 already hit on July 1, and if your list is above that number, sending is paused until you upgrade or delete contacts. But August 13, the automation cap, is what nobody’s talking about yet.

MailerLite framed these caps alongside “unlocks” like email templates, promotional pop-ups, and a second user seat, which is a bit like shrinking the rooms in your apartment and handing you nicer curtains for the ones that remain.

Why 3 automations is what breaks things

The subscriber cut gets the headlines. The automation cap is what breaks the funnel.

Walk through it. A reader downloads your book 1 reader magnet from the back of your ebook, which triggers an automation: deliver the magnet, wait two days, send a welcome email, wait three days, pitch book 2. That’s one automation, and four of your five allowed steps, gone on a single entry point.

Add a welcome sequence for people who subscribe through your website, with a different entry point and different content. That’s two.

A new release notification for your existing list when book 3 drops. That’s three. You’re done.

You haven’t built a reader magnet for book 2, there’s no re-engagement sequence for cold subscribers, and if you write under two pen names in two genres, you ran out of automations before you finished setting up the first name.

The five-step limit per automation is the other constraint, because the trigger doesn’t count but every email, time delay, and tagging action does. A welcome sequence that sends four emails with delays between them uses all five steps, which means there’s no room for a tagging action at the end to segment your list. Your automation works, but it can’t hand off to the next one in a way that keeps your data clean.

The free plan can’t support a functional email funnel for most indie authors anymore.

This is a pattern

MailerLite didn’t become the go-to recommendation for indie authors by accident. Five years ago the free plan was generous, with 1,000 subscribers, unlimited automations, and a clean interface that didn’t make you feel like you were fighting Salesforce every time you wanted to send an email. Every “how to build your email list” blog post for half a decade said the same thing: start with MailerLite.

That advice built a dependency. Authors set up funnels, connected reader magnets, linked back-of-book calls to action, and built sequences they stopped thinking about, infrastructure that hums in the background until someone changes the rules.

Mailchimp ran the same playbook before them, with a generous free tier to drive adoption and progressive cuts once the user base was locked in. The economics are not mysterious: free users are an acquisition cost when the platform is growing, and when growth slows, those users become an expense. The free tier shrinks until enough people convert to paid.

None of this is unusual, and MailerLite is a business adjusting its margins, doing what any company does when the economics shift. But the result is that the most recommended tool in indie publishing now offers a free plan that can’t support a new indie author’s needs, while the guides that made the recommendation haven’t updated.

What to do before August 13

Four options, and each one costs you something.

Stay on MailerLite, upgrade to Comfort. That’s $12/month for up to 500 subscribers, which gets you 50 automations, 10 forms, and 10 landing pages. A real tool at that tier. The question is whether $144/year makes sense for your revenue: if you earn $50/month from KDP, that’s a real percentage of your income; if you earn $500/month, it’s a rounding error. Only you know which side of that line you’re on.

Switch to Kit (formerly ConvertKit). The free plan covers up to 10,000 subscribers, with no send limits. Sequences, meaning linear email series for reader magnet delivery, welcome series, and new release announcements, are unlimited on the free tier. Visual automations are limited to one, but sequences handle what most indie authors need day to day. Migration from MailerLite takes an afternoon, and Kit publishes a step-by-step import guide.

Move to Substack. Free, with unlimited subscribers and no automation caps, but a different animal entirely: no reader magnet delivery, no back-of-book funnels, no tagging or segmentation. Substack is a newsletter platform, not an email marketing tool, and the distinction matters. If your list exists to support book sales, wrong fit. If it exists to build an audience you write to on a regular basis, it might be the right one.

Accept the limits. Consolidate down to 3 automations, delete the rest, trim your list to 250, and run a lean setup while you decide what comes next. Some authors sitting at 200 subscribers with two automations won’t feel August 13 at all, and if that’s you, this whole article is academic.

The advice in indie publishing has always been to start free and upgrade when you can afford to. Free KDP account, free Canva, free BookFunnel trial, free MailerLite. Build first, monetize later.

That’s decent advice when the free tiers are stable, but it stops working when they aren’t. The author who followed every guide, built their funnel on MailerLite Free, grew to 400 subscribers, and earns $200/month now faces a choice between $12/month for a tool that used to cost nothing, or spending a weekend migrating to a platform they’ve never used, while their back-of-book links still point to the old signup forms. Neither option is catastrophic, but both are friction nobody warned them about.

I’m not saying leave MailerLite, because the paid product is good and $12/month is not an unreasonable ask. What I’m saying is the recommendation pipeline, the YouTube videos, the Reddit threads, the blog posts that say “just start with MailerLite Free,” haven’t caught up to what “free” means in July 2026. If you’re still on that plan, you have three weeks to figure out what it means for you.

Subscribe now

I didn’t know what PSP stood for until this weekend.

I was scanning KDP news for this newsletter – the usual sweep of community forums, policy updates, the kind of reading that makes you feel like a tax accountant who accidentally ended up in publishing – and I kept seeing the same three letters. PSP. PSP deadline. PSP program. August 1.

I use Revolut. I publish from Poland. My first thought was: should I be worried?

I spent an hour digging. Revolut Bank UAB holds a full EU banking license from Lithuania, issued December 2021. In Amazon’s taxonomy, that makes it a deposit-taking bank. Not a PSP. I’m fine.

Then I thought about a subscriber of mine in Manila.

And an author I’d never met in a KDP Community thread, posting from Morocco: “The requirements for wire transfers are unrealistic and inappropriate for individual authors and small publishers.”

And another, in a different thread, who’d used the same bank account to receive royalties for ten years. Flagged out of nowhere: “My bank says they have no idea what this is about.”

Those authors have eleven days.
What a PSP is (and isn’t)

A Payment Service Provider is a company that isn’t a bank but gives you bank-like account details to receive international payments. Payoneer is one. PingPong. LianLian Global. Dozens of others, mostly serving authors outside the US and EU who don’t have access to the traditional banking infrastructure that Amazon’s systems were built around.

Amazon used to accept all of them. Now it doesn’t.

Starting August 1, 2026, KDP will only send royalties to two types of accounts: a deposit-taking bank (your local Chase, Barclays, PKO, or a licensed fintech like Revolut), or a PSP that has enrolled in Amazon’s Payment Service Provider Program and completed their KYC verification.

The participating list has about 30 companies on it. You can check it here.

If your payment provider isn’t on it and isn’t a licensed bank, your account goes on hold. Not just your payments. Your ability to publish. You can’t push new titles. You can’t update existing ones. Everything freezes until you switch to an eligible account.

Amazon started sending warning emails on June 15. If you received one and haven’t acted, the clock is running.

Who this hits

Not me. Not most of my American or European readers. If you have a checking account at a traditional bank, you will not notice this deadline pass.

It hits authors in countries where traditional banking infrastructure is thin and PSPs are the primary way to receive international payments. The Philippines. Morocco. Uganda. Bangladesh. Parts of South America. Countries where opening a USD-receiving bank account means navigating bureaucracy that was never designed for a solo author earning $400 a month in ebook royalties.

These are also the countries where self-publishing on KDP represents one of the most accessible paths to earning in dollars. The platform that removed geographic barriers to publishing is now enforcing geographic barriers to getting paid.

The PSP program exists for real reasons. KYC compliance. Sanctions screening. Anti-money-laundering. Fraud prevention. Amazon isn’t doing this to hurt authors. They’re doing it because financial regulators require it and because some PSPs weren’t meeting their obligations.

But intent and experience are different things. And the experience, for an author in the wrong country, is that a platform they’ve published on for years just told them to fix their banking in six weeks or lose access.

The confusion makes it worse

The deadline would be manageable if the system worked. It doesn’t.

KDP Community threads from the past two weeks show authors whose PSPs are on the participating list still receiving warning emails. One author’s bank has been flagged despite being a deposit-taking institution she’s used for a decade. Amazon’s support responses are generic – copy-pasted links to the same FAQ page the author already read before she contacted them.

Payoneer is on the participating list. It was also pulled for new accounts without any announcement. An author in Morocco discovered this after weeks of trying to set up a payment method, contacting both Payoneer and Amazon, getting contradictory answers from each.

If you get the warning email, you don’t know whether your PSP failed to enroll, failed Amazon’s KYC check, or whether the system just misfired. The penalty for guessing wrong is the same in all three cases.

What to do this week

Log into KDP. Go to Account Settings. Look for a yellow or red banner about your payment method.

If there’s no banner, you’re clear. Amazon has already classified your account as eligible. Move on with your week.

If there IS a banner, check the participating PSP list. If your provider is on it, contact KDP support and reference the list – your PSP may still be completing its KYC verification, which is their problem to solve, not yours, but you need to push them.

If your provider is NOT on the list, you have two options before August 1. Switch to a participating PSP – Payoneer is the most widely available one still enrolled. Or open an account at a deposit-taking bank that accepts international transfers.

In some countries, neither of those is simple. If you can’t get there in eleven days, contact KDP support now to document your situation. Pull any pending title updates or new submissions through before the freeze hits. And start the process of opening an eligible account so the hold is as short as it can be.

Full FAQ from Amazon here.

Amazon built the largest self-publishing platform on earth. It removed the gatekeepers for who could write and sell a book. Then it built a payment system that works best in the countries that needed that access least.

I don’t think it’s malicious. I think it’s unfinished.

For eleven days, the distance between those two words is someone else’s publishing career.

So I've been doing KDP work for a while now and I figured I'd just put this out there for anyone who needs it.

Last year I formatted and uploaded a 175,000-word non-fiction book. 650 pages. The client left a 5-star review and wrote "handled the project as if it were his own." That was a formatting and upload job. But the part I actually got obsessed with was the launch strategy I built around it – back cover copy, a full Amazon listing written for SEO, keyword and category research (picking categories where you can realistically hit #1, not just the obvious ones everyone piles into), pricing with royalty math for both print and Kindle, and two Facebook ad concepts with targeting angles.

Six pages of strategy for one book. That's the kind of thing that gets me going.

I also do AI-powered cover design. I spend real time researching comps in your genre and then design covers that actually look like they belong on the bestseller shelf next to them, not like someone typed "fantasy book cover" into Midjourney and called it a day. I've done cozy mysteries, sci-fi, noir thrillers, health non-fiction, literary family drama. If you can't afford a designer charging $500+ but you don't want something that screams "self-published" either, this is that middle ground.

What I can do for you, roughly:

If you just need formatting and upload – you send me a clean manuscript, I format it for KDP (paperback, hardcover, ebook, whatever combo you need), handle the metadata, get it uploaded and looking right. One round of revisions. For a standard-length book that's somewhere around $150-250. Longer books cost more because they take more time, and I'm not going to pretend a 200,000-word manuscript is the same job as an 80,000-word one.

If you want the full launch package – formatting plus keyword research, category selection, Amazon listing copy, back matter setup, cover coordination or AI cover design, and a couple rounds of revisions. That's $400-600 depending on scope.

If you want me to basically be your launch team – everything above plus I run the entire cover process (or design it myself with AI), build your pricing strategy, set up your first KDP ad campaign, and check in after 30 days to see what's working. $800-1,200.

Ground rules so nobody's surprised: 50% upfront. I need to see your manuscript before I quote anything because if it's a mess that needs cleanup before I can even format it, that's a different conversation. Page count matters – a 120-page memoir and a 700-page reference book are not the same project. And revision rounds are capped so we don't end up in an infinite loop where I'm reformatting the same chapter for the fourth time because you changed your mind about fonts.

DM me if any of this sounds like what you need. Happy to look at your manuscript and tell you honestly what it would take. If you let me write up your project as a case study on this Substack, I'll knock 10-15% off.