Founder knowledge engine
Arif KhanFree PDF — sent on signup
The AI Agent Operating Manual.
The exact stack of agents I use and the prompts that run them. From a founder running companies where humans and AI agents have real jobs — not demos, not theater.
Subscribe and the PDF lands in your inbox immediately. After that, weekly field notes from the same operating context.
What you get
The named agent roster
Every agent I trust with real jobs — what each one runs, where it lives, and which company it touches. Not a generic list of tool categories.
The exact prompts
The prompts I actually use, copy-pasteable. Not curated for show — the working set, with the mistakes that taught me to write them this way.
What broke and how I fixed it
Field notes on the failures: where ownership got blurry, which agents hallucinated, and the review loops I had to build to catch it.
Stay in the loop
Join the journey
Weekly notes on building companies with AI agents — what's working, what's not, and what I'm learning along the way.
Join 50+ founders
What's inside the manual
01
The named agent roster
Every agent I trust with real jobs — what each one runs, where it lives, and which company it touches. Not a generic list of tool categories.
02
The exact prompts
The prompts I actually use, copy-pasteable. Not curated for show — the working set, with the mistakes that taught me to write them this way.
03
What broke and how I fixed it
Field notes on the failures: where ownership got blurry, which agents hallucinated, and the review loops I had to build to catch it.
Who this is for
Founders and operators past the “should I use AI” question and into the “what should I actually hand off and how do I trust it” question. People with one or two agents already in production who want to design the next ten properly.
If you're still in demo mode, the manual will be premature. If you've already given an agent a real job and want to see how someone else has structured the same problem — open it.
Optional — The Nikita Test
Five questions to know if your agent stack is real.
Before you read the manual, answer these honestly. The ones you can't answer are the ones the manual is for.
- Can you name every agent in your stack and what it owns?
- Do you have written prompts for each, or are they trapped in chat windows?
- If an agent silently fails tomorrow, would you notice today?
- Is there one human accountable for each agent's output, or is it diffuse?
- Have you logged a single agent failure this week — and what changed because of it?
If you want to read first
The field notes are the long-form version of what ends up in the manual. Start Here is the curated reading order.