Ground Truth: Start Here

Most people use AI like a slot machine: pull the handle, hope something good falls out. This program turns it into a system that runs when you're not at the keyboard, and then into income. Every lesson works the same way, so learn the routine once.

Step 1: Open your AI program

Use one of these. Stay with the same one through a module when possible. The free tier of any of them is enough to finish Phase 2.

• ChatGPT: chat.openai.com
• Claude: claude.ai
• Google Gemini: gemini.google.com
• Microsoft Copilot: copilot.microsoft.com

Step 2: Read the lesson

Review the outcome and the explanation, then find the drill and the prompt at the bottom of the lesson.

Step 3: Copy the matching prompt

Each lesson has a prompt box. Press Copy prompt to copy it.

Step 4: Edit the [bracketed] placeholders first

If a prompt contains [bracketed text], paste it into a blank Word or Notes document first. Replace every bracketed placeholder with your own details, review the edited prompt, then paste it into your AI program.

Step 5: Continue the conversation

Ask follow-up questions, request examples, and tell the AI what needs improvement. The first answer is a draft, not a verdict.

Step 6: Complete, verify and save

Do the lesson drill, verify facts and calculations, and save useful prompts and finished work by module and lesson. Then press Mark as complete to track your progress.

How the program is organized

• Phase 1, Starter (free): Module 01.
• Phase 2, Foundations: Modules 02 to 05. Read, learn the basics and practice with simple prompts.
• Phase 3, Pro: Modules 06 to 16. Application work: you build real assets in your AI program, publish, automate and package what you learn.

Do the drills. Reading end to end and building nothing produces the same result as reading nothing.

Most people start an AI course with no answer to "what is this for," and that's why most people don't finish one. Learning a skill with no destination attached feels like homework by week three. So decide first, provisionally, and let it pull you through.

The skill you're about to build converts to value five ways:

• Leverage in your existing business or job. The most reliable and least glamorous. You produce more with the same hours. No new customers required, and it compounds quietly.
• Done-for-you services. Build content engines and follow-up systems for local businesses. Highest hourly rate, hardest to scale, easiest to start.
• Productized offers. A fixed package at a fixed price: "your follow-up system, built in a week." Easier to sell than hourly, easier to deliver repeatedly.
• Teaching. Workshops, cohorts, this curriculum. Requires you to be visibly two steps ahead, which, by the end of this program, you will be relative to most people.
• Referral and affiliate income. Legitimate when you recommend tools you actually use and disclose the relationship. Not a business by itself.

Pick one now. You are allowed to be wrong. The final module is where you commit properly, with better information. What matters today is that every lesson from here has somewhere to go.

Help me pick a provisional destination before I start this course.

MY SITUATION: [current work, hours available weekly, existing audience or network, financial runway]
WHAT I'M GOOD AT: [briefly]
WHAT I'M AVOIDING: [be honest]

Rank these five for MY situation, best first, with reasoning:
1. Leverage in my existing business
2. Done-for-you services
3. Productized offers
4. Teaching
5. Referral/affiliate

Then pick one and defend it. Name the trade-off I'm accepting and what I'm giving up by not choosing the runner-up.

Remember this is provisional. I'm choosing a direction to learn toward, not signing a contract. Tell me what would have to be true for me to change my mind later.

A language model is a prediction engine. It has read an enormous amount of text and learned, statistically, what tends to come next. When you type a question, it isn't looking anything up. It is producing the most plausible continuation of your words, one piece at a time.

That single fact explains almost everything you'll experience. It's why the model is superb at anything where "plausible" and "correct" are the same thing: rewriting, summarizing, brainstorming, structuring, translating tone, drafting from material you supply. It's why it's unreliable at anything where plausible and correct come apart: specific numbers, citations, quotes, what happened last Tuesday, what's in a document it hasn't seen.

Stop thinking of it as a search engine that talks. Think of it as an extremely well-read, fast, tireless assistant who will confidently guess rather than admit they don't know. You would never let that person send an email under your name unread. Same rule applies here.

I want to understand your limits before I rely on you.

In plain language, no jargon, answer these:
1. What kinds of tasks are you genuinely reliable at?
2. What kinds of tasks do you fail at in ways that are hard for me to notice?
3. What should I always verify myself before I publish or send it?

Be blunt. Do not reassure me.

The industry calls it "hallucination," which makes it sound like a rare glitch. It isn't. It's the same mechanism that makes the model useful, running on a question where the plausible answer and the true answer aren't the same thing. The model has no internal flag for "I don't know." It produces the shape of an answer because that's what it does.

Four categories cause almost all real-world damage:

• Numbers and statistics. Percentages, study results, market sizes. Invented figures look exactly like real ones.
• Citations and sources. It will produce a plausible author, a plausible journal, a plausible year, and a paper that doesn't exist.
• Quotes and attributions. Especially quotes from real, named people.
• Anything with a date on it. Prices, features, who holds a role, what a law currently says.

Draft this content for me, and then mark it up for fact-checking.

CONTENT: [paste or describe]

After the draft, add a section called VERIFY BEFORE PUBLISHING that lists every claim in the draft that is checkable (every number, date, name, quote, price, and factual assertion) as a numbered list.

For each one, tell me how confident you are (high / medium / low) and where I should go to confirm it.

If you are not confident in a fact, say so instead of smoothing over it.

Everything the model knows about you, right now, is in the conversation window. Your instructions, your pasted documents, its own previous replies: all of it is context, and context is finite. When a long conversation starts drifting, forgetting your rules or repeating itself, you have usually run past the useful edge of that window.

Three consequences worth internalizing:

• Long chats degrade. Start a fresh conversation for a new task rather than continuing a sprawling one. Bring forward a short summary instead of the whole history.
• Vague in, vague out, but stuffed-in also fails. Twelve pages of background where two paragraphs would do makes the important instruction harder for the model to weight.
• It cannot see anything you didn't give it. Not your website, not your calendar, not your last email, unless the tool explicitly has access and you've turned it on.

The skill this unlocks is the one in Module 06: deliberately assembling the right context once, and reusing it, instead of re-explaining yourself every morning.

Before we continue, produce a HANDOFF BRIEF I can paste into a fresh conversation so a new session picks up exactly where you are.

Include, under 250 words:
- The goal I'm working toward
- Decisions we've already made (so they don't get relitigated)
- My stated preferences and constraints
- What's done, and what the next step is
- Anything you'd warn a new session about

Write it addressed to the next assistant, not to me.

There are a handful of serious general-purpose assistants (ChatGPT, Claude, Gemini, Grok and others) plus specialists for images, video, presentations and code. Newcomers make one of two mistakes: they subscribe to five things, or they refuse to pay for any and then conclude AI "isn't that good."

Here is the honest guidance. The gap between the free tier of a good assistant and the paid tier of the same assistant is bigger than the gap between competitors. Longer memory, file uploads, persistent projects, image generation and better reasoning come with paying for one, not with collecting many.

So run everything free through Module 06's first lessons. By then you'll know which one's writing you actually like, because they do have different default voices. Pay for that one. Add a specialist only when your primary has visibly failed at a job you need done repeatedly.

One caution on brand loyalty: rankings churn constantly and the "best model" arguments online are mostly noise. Pick on fit and stop reading the leaderboards. Verify current prices on each tool's own pricing page before you buy.

I'm choosing a primary AI assistant. Interview me first.

Ask me 6 questions, one at a time, about: the work I do, what I'd use this for weekly, whether I handle sensitive information, how much I write vs. analyze vs. create visuals, my budget, and my technical comfort.

Wait for each answer before asking the next.

Then recommend ONE assistant as my primary and explain the trade-off I'm accepting by choosing it. Name what it's worse at than the alternatives.