← All free guides
Free guide

The One-Day AI Lesson Plan and the Prompt That Builds Each Page

Most people try to learn AI from a long video or a list of terms, and none of it sticks. This is a one-day plan with six ideas, one per time slot, and one Claude Code prompt that builds you a one-page lesson on each idea: plain words, one story, one picture and a 3-question check.

A free guide by Zach Crawford

Most people try to learn AI from a long video or a list of terms. The words go in, and by the next day they're gone, because there's nothing to hang them on.

Here's how I'd learn the basics of AI in one day. One idea per time slot. For each idea, Claude Code builds you a one-page lesson with plain words, one story and one picture, and 3 questions at the end so you know it stuck.

One idea, one page, one story.

The one-day plan

The gaps are on purpose. Take lunch, and give the morning a chance to sink in.

Step 1. Ask Claude Code for one page on each idea

Make a new folder on your computer called AI Day. Open Claude Code in that folder.

At each time slot, start a new chat, paste the prompt below, and fill in the three lines at the top. The idea line for each slot is in the steps below. Fill in your business once and keep a copy, so you only type the idea line each time.

Why a new chat each time? You'll know by 9:30.

Build me a one-page lesson on one AI idea. One idea, one page, one story.

The idea: [the idea for this time slot]
My business: [what you sell, and who buys it]
How I use AI now: [a line, or "not much yet"]

Make it one HTML file in this folder, named after the idea, like lesson-tokens.html. When it's done, open it in my browser.

Put these on the page, in this order:
1. What it is. Plain words, for a smart person who has never written code. Short sentences. If you need a technical word, explain it the first time you use it.
2. One story that makes it click. Use an everyday thing anyone knows, like a kitchen, a library or a map. Write it as an example ("Picture a..."), not as something that happened to me or to anyone real.
3. One simple picture or diagram that shows how it works. Draw it inside the page with code, no image files. Label each part in plain words. Keep the labels as big as the body text, including on a phone. On a narrow screen, stack the parts top to bottom instead of shrinking the whole picture.
4. Why it matters for my business. Use only what I told you about my business. Show where this idea comes up when I use AI in my work. If you need a fact about my business that I didn't give you, ask me before you build. Don't make up stories, numbers, results or quotes about me, my business or my customers.
5. A 3-question check at the end. Multiple choice. Hide each answer until I click, then show the right answer and one line on why.

Rules:
- Every fact on the page must be true. If you're not sure of something, leave it out. Don't say "always" or "never" unless it's true every time. Don't add numbers, dates, prices or product names unless you're sure they're right today.
- No outside services. Everything goes inside this one file: no web fonts, no images or scripts from the internet, no links out. It has to work with the internet off.
- Keep it short enough to read in one sitting. Big, easy-to-read text, plenty of space, and it has to look right on a phone too.

When it's done, tell me the file name and the one sentence I should remember.

If Claude Code asks to open the file, say yes. If it doesn't, double-click the file in your AI Day folder. It opens in your browser.

Got a question wrong in the check? Don't skip it. Paste this in the same chat:

I got question [number] wrong. Explain that part another way, with a different everyday example, and add it to the page above the check.

Step 2. 9 AM: Tokens

In plain words: AI reads text in small pieces called tokens, like Lego bricks. A token can be a whole short word or a piece of a longer one.

Why it matters for your business: Everything in a chat counts in tokens: your messages, its replies and every file it reads. There's a limit to how much it can hold at once. That's why a long chat fills up. When Claude Code gets close to full, it squeezes the older part into a summary, and details from early on can get lost. So start a new chat for each new job, and put the rules you always want in a file called CLAUDE.md. Claude Code reads it at the start of every chat.

The idea line:

Tokens, and why a long chat fills up

Step 3. 10 AM: Prediction

In plain words: AI writes one token at a time. Each time, it picks a likely next piece based on everything that came before it.

Why it matters for your business: It's built to write what's likely, not what's true about you. Left alone, you get a likely answer, and likely often means generic. Give it your real facts, your examples and your own words, and the likely next words start to sound like you.

The idea line:

Prediction: how AI writes one piece at a time

Step 4. 11 AM: Embeddings

In plain words: AI puts every word on a map. Words that mean the same thing sit close together, like "cheap" and "low price." Words that mean different things sit far apart.

Why it matters for your business: Search by meaning runs on this. It's how my own course library works. Every course I've bought is in there. I ask a question in plain English, and it finds the lesson that means the same thing, even when the words are different.

The idea line:

Embeddings: how AI turns words into spots on a map

Step 5. 1 PM: Hallucinations

In plain words: Sometimes AI says something that sounds right and is wrong. That's called a hallucination. It goes back to 10 AM: it writes what's likely, and a likely sentence can still be false.

Why it matters for your business: Never post a number, quote, price or fact from AI without checking it. Tell it it's OK to say "I don't know." Ask it where each fact came from. And give it your own files, which is the next slot.

The idea line:

Hallucinations: why AI sometimes makes things up

Step 6. 2 PM: RAG

In plain words: RAG stands for retrieval augmented generation. AI doesn't know your business. RAG is an open-book test: give it your files, and it looks up the right page before it answers.

Why it matters for your business: Your offer, your prices and the questions your students ask live in your files, not in the AI. When it pulls the right part of your files first, the answer comes from your business instead of a guess. A help bot that answers student questions from your own lessons is a common use.

The idea line:

RAG: how AI finds the right part of my files before it answers

Step 7. 4 PM: Agents

In plain words: A chatbot answers. An agent does the job, like a new hire with a to-do list: it reads, does the work, then checks it.

Why it matters for your business: You hand it a job, not a question. My agents do the research and hand me a brief. Claude Code works this way too. You give it a job, it gathers what it needs, does the work and checks it. You still check what it hands back.

The idea line:

Agents: how AI takes steps instead of just answering

The limits, honestly

Want the next guide? Follow @iamzachcrawford