AI systems that work while you sleep.

Chapter 1: Twenty Minutes to a Working System

I know why you're here. You want to make content.

You've seen the accounts pumping out three videos a day: faceless channels with AI narration, stock footage, and titles like "5 AI Tools That Will Replace Your Job in 2026." Some of them are making real money. You want in.

So you opened ChatGPT. You typed "write me a YouTube script about AI productivity tools." You got something back. It was... fine. Generic. The same script anyone else would get from the same prompt. You posted it anyway. Twelve views. You tried again the next day. Started over from scratch. Re-explained your niche. Re-described your style. Got back something that sounded like a different person wrote it. Posted that too. Nine views.

You're not doing anything wrong. You're doing what everyone does. And that's the problem.

You're treating AI like a vending machine. Put a prompt in, get a script out. But a vending machine gives everyone the same snack. That's why all the content sounds the same.

This chapter fixes that. Not with a better prompt. With a system.

By the end of this chapter, you'll have:

• A running content bot that finds trending topics, generates video scripts, and delivers them to your phone
• A video pipeline that turns any script into a finished MP4, hands off, for free
• A system that remembers what it already covered, so no duplicate topics, ever

The workflow: Start your workspace → Pick your niche → Run the content scout → Connect Telegram → Make a video

Step One: Start Your Workspace

When you buy the book, you get an email with a link and a setup code. The link looks something like your-name.zero2ai.dev. You paste the code in, hit go, and your system starts building in the background.

That's the only setup in this chapter. Nothing to install, no API keys to wire up, no config file to hand-edit. It takes about ten minutes, and there's nothing to watch while it runs. So the chapter puts those ten minutes to work on the one part of the build no software can do for you.

Prefer to run everything on your own machine instead? Appendix H has the self-hosted path. Same system, you maintain it yourself.

Step Two: Pick Your Niche (Do This While It Builds)

Your workspace ships with a content skill already loaded. It's called daily-content, and it does four things:

1. Finds what's trending in a niche
2. Writes short-form video scripts on those trends
3. Stores every script and topic so it never repeats itself
4. Hands you five scripts a day, ready to record

No editing. No posting. No thumbnails. Just the scripts. It arrives pointed at consumer tech and AI tools, and about fifteen minutes later you point it at yours.

So you decide now. Two things:

One: your niche.Personal finance, real estate, fitness, gaming, cooking, small business, trades. Whatever you can talk about every day without getting bored. Be specific. "Fitness" is a category. "Home workouts for people with bad knees" is a niche, and a niche is what a trend search can actually feed on.

Two: one line your system never crosses.Every channel has one. If your niche touches money, health, or law, this one isn't optional: your system never presents itself as a licensed anything. No "financial advisor," no "coach," no "we recommend" in a voice that sounds like a credential. YouTube pays for a creator with a point of view. It doesn't pay for an AI persona dispensing expert advice on a sensitive subject, and it says so by name.

You can absolutely make money content. You make it as yourself talking about money, not as a synthetic expert.

Your First Run

The daily-contentskill is sitting there waiting. It loads automatically on startup, which means you don't have to turn anything on. You click Chat in the sidebar, type this, and press Enter:

make content

That's it. That's the whole command.

The first run takes 30 to 60 seconds. The system scans trends, checks its topic list (empty on the first run, so everything is new), generates scripts, and stores results. You watch it work in real time. What comes back is not just a script. It's a complete short-form video package:

SCRIPT 1: AI Video Just Became an API Call

HOOK (0-3s):
You are paying a monthly seat for something that bills by
the clip.

SCRIPT:
[0-10s] Video generation used to mean a subscription and a
credit balance. Now the model providers expose it the same way
they expose text: you send a prompt, you get back a file.

[10-25s] The pricing changes shape with it. Instead of a flat
monthly fee whether you render one clip or fifty, cloud video
generation runs roughly five cents to fifty cents per clip,
depending on the model and the length.

[25-45s] That flips the math for anyone posting less than daily.
Count how many clips you actually rendered last month. Multiply
by fifty cents. If that number is smaller than your subscription,
you are paying for the idle days.

ON-SCREEN TEXT:
- Video generation is now a per-call API, not just a subscription
- Roughly $0.05 to $0.50 per clip
- Count last month's renders before you renew

HASHTAGS: #AITools #ConsumerTech #AIVideo #CreatorTools
SOURCES: openrouter.ai/docs (linked)

Timed script beats you can read straight into a teleprompter. On-screen text overlays for your editor. A caption ready to paste. Hashtags. Source links, so you're not making things up.

You asked for a content scout. You got a content production line.

The system found five topics like this, each sourced from real trending data: release notes, pricing pages, developer changelogs, launch posts, teardown threads. Not "top 10 AI tools" recycled from 2023. Actual changes from this week, turned into scripts you can record today.

Read all five and notice something else. They don't share a shape. One opens on a price, one opens on a broken assumption, one walks a comparison. That isn't luck. There's an originality rule in your skill: a batch has to vary in structure, not just in subject, and each script has to carry a claim its source doesn't make. The book comes back to why that rule is the difference between getting paid and not.

Is it slop? Yeah, mostly. But it's slop with real data, real sources, and a structure that's ready to shoot. That's already better than what most people are posting.

It worked. That's the magic moment, and you didn't configure a thing to get it.