Your job is the judgment. The AI writes the code and runs the ads. You write down what only you know.

That is the knowledge: your methodology, your rules, and the answers that make your service yours. Per the site, it gets designed once, and every AI worker runs against it. This is the real work; everything else is execution.

What to write down:

1. What you sell, in one plain sentence. Who it is for and what it costs.
2. Your method. The steps you follow and the order you follow them in.
3. Your rules. What you always do, what you never do, and where you say no.
4. Your voice. Real examples of how you talk to customers, so the ads and follow-ups sound like you.
5. Your answers. The questions buyers ask before they pay, and your response to each one.
6. Your line. When a refund, exception, or edge case needs a human call, that call is yours.

The kit itself is free. The 206-page PDF or the AI crawl token costs nothing. The paid path is the 3 week program and the community, and there is a paid product like Prigmar if you want packaged AI employees. No team, no funding, no permission needed. Start with the free kit.

Mouldi

Divide revenue by hours worked. That's it. If you made $4,000 in month three and spent 60 hours, your rate is $66.67 an hour. Total revenue tells you nothing about whether the business is worth your evenings.

Here's the loop I use.

1. Log every hour you touch the business. Building, marketing, support, invoices. One number per week.
2. Log revenue collected, not booked. Cash that landed is the only revenue that counts.
3. Divide. Revenue collected divided by hours logged. That's your real rate for the period.
4. Track it monthly for 12 months on one row of a spreadsheet.
5. Compare months. Early months are ugly. Weekends of setup with no sales means a low rate. That's normal, not failure.
6. Watch the trend, not the first number. If month six pays more per hour than month two with the same effort, the automation is working.

The 12 systems in the Free Builder kit cover the whole loop. The free version is a 206-page PDF or an AI crawl token, no card. Paid options are the $19 AI employees guide and the 3 week program. The number you want gets honest around month four, once the early build hours stop repeating.

Record it. Adjust. Keep going.

Mouldi

You write one document before the AI worker starts. That is the knowledge file: your methodology, your rules, and the answers that make your service yours. Everything else gets generated in the loop.

Here is the order that keeps you from writing too much or too little.

1. Name the one thing you sell and who buys it. One paragraph.
2. Write the rules your work follows. The non-negotiables. Ten to twenty bullets is common.
3. Write the answers to your five most asked customer questions, in your own words.
4. Write three examples of work you would sign your name to. Real ones, not descriptions of them.
5. Stop there. Take it to the AI worker and let it start. Add to the file as gaps show up.

That file is the real work. The 206-page PDF and the AI crawl token do not replace it and they do not write it for you. They give your worker the architecture, the module spec, and the scaffold so your context lands in something coherent instead of a pile of one-off prompts.

The kit is free. The AI crawl token fetches every chapter, module spec and file straight from the server so your worker reads it while it works. The 3 week program and Prigmar are paid, $19 for the AI employees guide.

Keep the file short at the start. You will edit it more than you write it.

Mouldi

The token is an access credential, not a meter, so an AI worker crawling the kit chapters fetches what it asks for. Nothing in the kit spec says the token counts fetches or expires after a set number of them. If you want to know what your worker read, ask it to list the files it pulled, or have it write a short summary per chapter after each fetch. That record lives in your worker's output, not in a dashboard on my side.

If a run goes sideways, stop it. You control the worker, so kill the process or revoke the token from your side and restart with a narrower instruction.

A tight loop:

1. Give the worker a short task that names the chapters and module specs it needs.
2. Have it log each URL it fetches, one line per fetch, with the file name.
3. After the run, read the log and compare against the 12 modules.
4. If it wandered, narrow the prompt to the exact files and run again.

What is free: the 206-page PDF, the reference codebase, the AI crawl token and prompt. What is paid: the 3 week program, the $19 Prigmar prompt set, and community tools when you choose them. I do not have a published fetch limit for the token yet. When I do, it goes on the site.

Keep the loop small. Check the log. Move on.

Mouldi