Before You Touch a Single Tool, Read This

Rich made this video with AI in just a few minutes. Imagine what you'll build in 30 days.

Skip ahead to Setup →

Most people think they know what AI is.

They've used ChatGPT. They've typed a question, gotten an answer, maybe felt a little impressed, maybe a little underwhelmed. And they've filed it under "useful sometimes" and moved on with their life.

If that's you — good. You're in the right place. Because what you're about to learn has almost nothing to do with what you've experienced so far.

I'm going to share three distinctions with you. They're not complicated. But they will fundamentally change how you think about AI, about the tools we're going to use, and about what's actually possible when you stop chatting with AI and start building with it.

These three ideas are the reason this program exists. Get them, and everything that follows will make sense. Miss them, and you'll spend the next few weeks wondering why we're not just using ChatGPT like everyone else.

Here's What Most People Get Wrong About AI

Let me give you a number that should stop you in your tracks.

Now, ask yourself this: What is happening in the gap between the 88% and the 6%?

Because the 88% are not stupid. They've bought the tools. They've run the pilots. They've sent their people to prompt engineering workshops. And they're still stuck.

The answer is that the 88% are doing the wrong thing with AI. They're chatting with it when they should be building with it. They're optimizing the wrong layer. And they're using tools designed for conversation when they need tools designed for construction.

That's what the three distinctions below are about. Not which AI is "better." Not which tool has more features. But which mode you're operating in — because mode determines everything.

What happens the day AFTER you get the answer?

With a chat interface, you type a question, get a response, close the tab, and tomorrow you start from scratch. Every conversation is a fresh start. There's no memory of what you built yesterday. No ability to save the instructions that worked. No way to hand off what you learned to a system that keeps running without you.

A coding CLI is different. It lives on your machine. It reads your files. It edits your documents. It remembers your project context across sessions. When you build something today, it exists tomorrow. And the day after that. And the day after that.

Imagine you need to get across town every morning. A chat interface is like asking a stranger for directions each day. You get a fine answer. Maybe even a great answer. But tomorrow morning, you're asking again. A coding CLI is like building a road. The first day takes longer. But from then on, you just drive.

One approach is tool-mode: bring a task to the AI, get an answer, leave. The other is infrastructure-mode: build persistent systems that compound over time.

The entire AI industry has been selling you on something called "prompt engineering" — the art of crafting clever instructions to get better answers. And prompt engineering isn't useless. But research from some of the smartest people in AI has broken down where results actually come from:

• 15% comes from which AI model you're using
• 10% comes from the prompt — your clever instructions
• 75% comes from everything else: the context you provide, the architecture around the AI, the system design, the data, the configuration

Most AI training teaches you to optimize the 10%. The prompt. The clever question. And they ignore the 75% that actually determines whether you get real results or just impressive-sounding responses.

This program teaches the 75%.

Let me tell you how I learned this the hard way.

I was a master power user of Evernote. Loved it. Used it for everything. And then I started working seriously with AI and hit a wall almost immediately. I couldn't get my notes out of Evernote and into AI easily. I couldn't get AI's output back into Evernote easily. Every interaction required me to manually drag files in, copy results out, paste them somewhere. The tool I'd relied on for years was suddenly a bottleneck.

So I moved to Notion, hoping it would be better. And it was — a little. Notion had an integration that let AI pull files and work on them directly. But it didn't always work. Sometimes the connection failed. Sometimes files didn't sync correctly. And here's the part that really burned me: when it didn't work, I was dead in the water. There was no copy of my files on my computer. Everything lived in Notion's cloud. If the integration broke, I had nothing to work with.

I stumbled on Obsidian almost by accident. And if I'm being honest, I didn't even like it at first.

But the more I used Obsidian with AI, the more I fell in love with it.

An Obsidian vault is just a folder on your computer. Plain text files sitting on your hard drive. You own those files. They'll still be readable in fifty years.

And because those files are plain text on your computer, your coding CLI reads them directly. Your notes, your documents, your project files, your business context — the AI sees all of it, instantly, without any export, conversion, or API. Your vault IS the context.

I think one of the major reasons I'm so far ahead of most people with AI is simply that I stumbled on Obsidian early. It gave my AI long-term permanent memory before anyone else had figured that out.

We've gotten really good at capturing and organizing information. We've built beautiful digital warehouses. But we're treating these systems like warehouses when we should be treating them like forges.

— Eva Keiffenheim

Google Docs, Notion, Evernote — they're warehouses. You store information there. It sits.

Obsidian — connected to a coding CLI — is a forge. Your knowledge goes in, and the AI transforms it.

The shift is not from one note-taking app to another. It's from storing knowledge to using knowledge.

Think about how most people use AI right now. They open a chat interface. They need AI to work on a document, so they go find the document, drag it into the chat, wait for the response, copy the output, go put it where it belongs, come back, drag in the next file, repeat.

The AI sits in one place. The human runs all over.

The AI does the thinking. The human does the fetching, the carrying, the filing, the organizing. The human is the assistant to the AI.

Now think about how it works with a coding CLI. You sit at your computer. You tell the AI what you need. The AI goes and grabs the files. It reads them. It works on them. It creates new files, organizes them, puts them where they belong, and tells you when it's done. You bark the orders. The AI does the running.

You take back your power.

That's not a small difference. That's a complete inversion of the relationship. In one mode, you serve the AI. In the other, the AI serves you.

Picture two people. Both are using AI. Both are smart. Both work hard.

• Person A asks AI for answers. Each interaction is independent. Nothing carries forward. Every day starts from zero.
• Person B builds a system that answers. Each session builds on the last. The system gets smarter every time they use it.

Person A's output is linear. Double the time, double the output. Stop working, output stops.

Person B's output compounds. The hundredth session is ten times faster than the first, because the system already knows everything.

Here's a real example. A guy named Alex McFarland spent about an hour building context profiles for his AI system. Then he rebranded his entire business in two days. Fifteen-plus pieces of professional content. Not because he wrote better prompts. Because he built a system that already knew him.

One hour of building replaced weeks of asking for answers.

The 88% are asking for answers. The 6% have built systems that answer.

This program teaches you to build the system.

What This Means for You

I built this system myself. I use it every day. The AI tools you're about to set up are the same ones running in my business right now. This is not theory I read in a book and decided to teach. It's infrastructure I built, tested, broke, rebuilt, and now rely on.

I'm not going to pretend the setup is effortless. There's a learning curve. You'll hit moments where you're frustrated. Here's what I've learned about those moments: you cannot learn unless you push through them. The frustration IS the learning.

A study of 667 people found that working effectively with AI is a separate skill from being good at your job. Years of experience, advanced degrees, deep expertise — none of it predicted who would get the best results from AI.

The difference was not intelligence. It was not domain expertise. It was the skill of building systems that let AI work with what you know.

That's the skill this program teaches.