Why I cut my AI book from 17 chapters to 11 before I finished it

Editor’s note (June 2026): Since I wrote this, the book got its final title — The Real AI Side Hustle — and it’s now live on Amazon in Kindle and paperback. Free reader resources live at charlesmcquain.dev/books/ai-income. The original post follows, unchanged.

I’m writing a book in public — working title AI Income, Field Tested — about using AI tooling to build a real side-income asset. A few weeks ago the outline was 17 chapters and the target was around 80,000 words, somewhere north of 290 pages. This week I cut it to 11 chapters across 4 parts, and the new target is a tight ~120–160 pages — roughly 45–52K words.

I did that before I finished drafting it. I was five chapters in. That feels backwards — you’re supposed to finish the thing, then edit it down — so here’s the honest reasoning, because the why is more useful than the decision.

The research that flipped the plan

I spent a week doing competitive research on the category before committing to a length. The pattern was hard to miss once I looked:

• The books people actually finish and recommend in this niche are short — roughly 100 to 150 pages. The reviews praise the brevity. “No fluff” is the highest compliment a how-to book gets.
• The padded “bestsellers” — the 300-page ones — collect the one-star reviews. And the complaint is almost always the same: this could have been a blog post, 50 pages of value stretched to 300, get to the point.

I had been optimizing for the wrong number. A bigger word count felt like more value. It isn’t. In a practical, do-this-then-that book, length is mostly a tax you charge the reader. So the target reversed: from “how much can I write” to “how little can I get away with while still teaching the whole method.”

Eleven chapters is what’s left when every chapter has to earn its place. Six of them are essentially the same content I already had, condensed. The other six chapters I’d outlined turned out to be padding wearing a chapter’s clothes — examples that repeated a point already made, a “history of AI” detour nobody needs, that kind of thing. Cutting them didn’t lose anything. It just removed the tax.

What “concise” actually costs to produce

Here’s the part that surprised me: cutting is more work than writing long, not less. Anyone can hit 80,000 words. Hitting 45,000 that are all load-bearing means going through finished prose and deleting sentences you were proud of.

One concrete example. Chapter 2 came in around 6,500 words on the first pass. I ran it back through a hard condensing edit and it landed at about 3,750 — and it got better. Same argument, same examples, none of the connective throat-clearing between them. Forty percent lighter and it reads faster and hits harder. When that kept happening chapter after chapter, the decision made itself.

The positioning was the real reason

The length change was downstream of a bigger realization, and this is the one I actually want to write down.

When I started, I thought my edge was honesty. The plan was basically: the side-income space is full of hype and fake screenshots, so I’ll just be the honest one. Real numbers, real failures, no course-selling.

That’s not an edge anymore. Everyone in this category now claims to be the honest one. “No hype, just real results” is the most common hype there is. Honesty isn’t a differentiator; it’s the floor. If you’re not honest you’re disqualified, but being honest doesn’t make you interesting.

So what’s actually left when honesty is table stakes? Two things, and they’re the only things a competitor can’t copy off me:

1. Proof, not claims. Real, dated, redacted dashboards from projects that are actually live. Not “trust me, it works” — here’s the screenshot, here’s the date, here’s the number with the sensitive parts blacked out. You can fake a claim. It’s much harder to fake a year of dated screenshots from real properties.
2. The AI-search shift. The way people find content is changing — answer engines and AI citations are starting to matter as much as the classic blue links, and almost nobody writing in this category is building for it yet. I am, in the open, and that’s a story I can tell because I’m living it.

Once I had that straight, the editorial test for every paragraph became brutal and simple: could a competitor write this exact paragraph without my projects? If yes, it’s generic advice that’s already free on a hundred blogs — cut it. If it can only exist because I built the thing and have the receipts, keep it. A shocking amount of the original outline failed that test. That’s the other half of why the book got shorter.

Writing it with AI without producing slop

Fair question, since the book is partly about AI tooling: how do I use AI to write it without it turning into the exact AI slop I’m telling people to avoid?

The method that works for me inverts the usual order. The AI doesn’t write and then I edit. The AI interviews me first — it asks what actually happened, what the numbers were, what broke, what I’d do differently — and only once it has the lived specifics does it help shape them into prose. I supply the story and the receipts; it handles structure and pacing. The judgment, the experience, and the facts are all mine. The result reads like me because the substance is me — the AI is a writing tool, not the author.

Slop happens when you ask a model to generate experience it doesn’t have. If you feed it real experience and make it do the formatting, you get leverage without the tell.

The build-in-public part

In the spirit of actually building this in public: as I write this, 5 of the 11 chapters are drafted, and the other 6 are scaffolded — outlined, with the key beats and examples noted, but not written. I’m targeting an August 2026 launch. That’s the real state of it, mid-flight, not a finished-product announcement.

I’d rather show the messy middle than a clean reveal at the end. Cutting a book in half before you finish it is exactly the kind of decision that gets sanded out of the polished launch story, and it’s the part I’d have found most useful to read from someone else.

You can see where the book stands — and get on the launch list — on the book page. And if you want the cut chapters’ worth of lessons as they happen, the journey is going out in the newsletter below.