AI, Explained
What AI actually is, which tools matter, and the one skill — prompting — that works in all of them. No math, no hype.
From mystery to method
Six short lessons, one useful skill, and a starter kit you keep. The final test checks ideas, not memorized wording. Tap any lesson to start it, or begin from the top.
What AI actually is
AI is a prediction machine: given everything it has seen before, it guesses the most likely next thing — billions of times — until a full answer appears. That is why it sounds fluent and why it can still be wrong.
- What AI is actually doing when it answers you
- Why a confident, fluent answer can still be wrong
- A simple way to think about AI that makes the rest of this course click
🎙 Nina explains · Read transcript
Artificial intelligence is a prediction machine. Given everything it has seen before, it guesses the most likely next word, pixel, or note — and it does that billions of times until a full answer appears. That is why it sounds so fluent, and why it can still be wrong. The useful mental model is not a brain in a box. It is a very fast guesser trained on most of the internet. Once you see that, everything else in this course clicks into place.
Prediction, not magic
AI doesn't look up the right answer or understand the way you do. It predicts likely text, pixels, or notes — over and over — until a full response appears.
The useful mental model
Think "very fast guesser trained on most of the internet," not "brain in a box." This one idea makes every later lesson click.
Two quick questions. Answers check instantly — no grade, just practice.
1. When AI answers you, what is it actually doing?
It's a very fast guesser, not a lookup machine. That's why it can sound completely fluent and still be wrong.
2. Why can an AI answer sound confident and still be wrong?
The machine guesses what sounds likely, not what is true. Confidence is not accuracy — that's the whole reason for lesson 4.
Explain it in one sentence
This locks the mental model in your own words. What comes back is one sentence you actually believe — and it becomes page one of your starter kit in lesson 6.
Meet the tools
ChatGPT, Claude, Gemini, Copilot, Grok, and Muse are different apps built on the same idea. Pick one to start — and see why the same prompt gets different answers in each.
- Which AI tools are worth knowing: ChatGPT, Claude, Gemini, Copilot, Grok, and Muse
- Why the same prompt gets different answers in each one
- How to set up one free account — any of them works for this course
🎙 Theo explains · Read transcript
ChatGPT, Claude, Gemini, Copilot, Grok, and Muse are different products built on the same idea — a prediction engine in its own app, with its own personality and strengths. In this lesson, create one free account: pick any of them. Then try the bonus exercise and watch the same prompt get different answers. That difference teaches you more than any review could — and it shows why your prompting skill transfers everywhere.
Same engine, different packaging
Each one wraps a prediction engine in its own app, personality, and strengths. The skill you build here transfers to all of them.
Free is enough to start
Every tool on this page has a free tier. That's all you need for this course — no card, no commitment.
We asked them all the same question
We ran one prompt through all six tools. Same words in, six different answers out. This is the prompt — it's what a person typed:
Dawn spills on the sea
A lighthouse wakes with the sun
Waves whisper good morning
Pale light meets the sea,
the beam yields to rising gold—
night's watch is complete.
Golden morning light,
Guiding ships through fading mist,
Silent beacon shines.
Dawn paints the still sea,
A lighthouse wakes with gold light,
Gulls greet morning's hush.
Beacon cuts the mist
Golden rays kiss the tower
Waves whisper awake
Beacon holds its breath
as dawn pours gold on dark waves —
light turns toward home
That's the whole lesson in one picture: the same prompt, six personalities. The skill you're building — saying what you want, clearly — transfers everywhere. The flavor changes with the tool.
Create your free account
Pick one of these to start — any one works for this course:
Two quick questions. Answers check instantly — no grade, just practice.
1. ChatGPT, Claude, Gemini, Copilot, Grok, and Muse are…
Each one wraps a prediction engine in its own app, with its own personality and strengths. Same idea, different packaging.
2. To take this course, what do you need?
Every tool on the list has a free tier, and the skill you're building works in all of them. One account is plenty.
Bonus: see it with your own eyes
This is where the lesson clicks — optional, but worth five minutes. Pick one tool from the list above (any one). Paste the prompt below into it, then change one word and run it again. What you're looking for: the same prompt, slightly different answers. That's the "different packaging" from the cards above. If you save your notes, they become the tool picks in your lesson 6 starter kit.
The prompting playbook
One skill works in every AI tool: say the outcome, give context, set constraints, name the format. Then iterate.
- The four ingredients of a prompt that works: what you want, the background, the limits, the shape of the answer
- What to do when the first answer isn't right: say what to keep, what to change, and what "good" looks like
- How to put it all together in one prompt you can reuse anywhere
🎙 Nina explains · Read transcript
Prompting is the one skill that works in every AI tool. A strong prompt has four parts: the outcome you want, the context it should know, the constraints to respect, and the format you want back. For example: plan three weeknight dinners for two, under sixty dollars, thirty minutes of cooking each, and give me a dinner plan plus a grocery list. Then iterate — tell it what to keep, what to change, and what good looks like.
Four parts of a strong prompt
Put it together
"Teach me the basics of sourdough baking. I have a free Saturday afternoon, a standard home oven, and zero experience. Explain it like I am a smart friend: a step-by-step plan plus a shopping list."
Two quick questions. Answers check instantly — no grade, just practice.
1. A strong prompt has four parts. Which set is right?
Say what you want, give the background, set the limits, and name the shape of the answer. That recipe works in every tool.
2. The first answer isn't quite right. What now?
Prompting is a conversation. Tell it what to keep, what to change, and what good looks like — the second draft is usually the good one.
Run the four-part prompt
This shows the four parts working together in one prompt. Run it in any tool from lesson 2. What comes back is a step-by-step plan plus a shopping list. Then improve it twice — say what to keep, what to change, what good looks like — and keep the final version for your starter kit.
Don't trust it blindly
AI invents facts with total confidence — engineers call it hallucination. Verify anything you will act on.
- Why AI sometimes invents facts that sound completely real
- A simple habit: what to double-check before you act on an answer
- How to write your own one-line safety rule
🎙 Theo explains · Read transcript
AI can be confidently wrong. It invents facts, quotes, and citations that sound perfectly real — engineers call this hallucination. It happens because the machine predicts likely text, not true text. So verify anything that matters: names, dates, prices, instructions you will act on. Ask it to show its sources, then actually open them. In this lesson you will catch it inventing something on purpose, so you learn what the failure feels like before it costs you anything.
Hallucination: confident and wrong
It invents facts, quotes, and citations that sound perfectly real — because it predicts likely text, not true text.
The verification habit
Check names, dates, prices, and anything you'll act on. Ask for sources, then actually open them.
Two quick questions. Answers check instantly — no grade, just practice.
1. What does "hallucination" mean for AI?
It predicts likely text, not true text — so the invented facts sound perfectly real. That's what makes it dangerous.
2. You need a real price for something you're about to buy. The AI answered confidently. You should…
Verify anything you'll act on: names, dates, prices. Ask for sources — then actually open them.
Catch it making something up
This teaches you what AI failure feels like before it costs you anything. Run it in any tool: ask for an obscure fact you can verify, or ask for sources and actually open them. What you catch becomes your written verification rule in lesson 6.
Beyond chat
Images, video, voice, music — the same prompting skill pointed at new outputs. This is the doorway to AI filmmaking.
- How the prompting you've learned works for images, video, voice, and music too
- Why every AI film starts with a clear description of what you want to see
🎙 Nina explains · Read transcript
Everything you have learned transfers beyond chat. Image generators, video tools, voice apps, and music makers all run on the same deal: you describe the outcome, the machine predicts its way there. The four-part prompt still applies — outcome, context, constraints, format — just pointed at pictures or sound instead of words. This is also the doorway to the film school: every AI film you will ever make starts exactly this way, with a clear description of what you want to see.
One skill, many canvases
Image, video, voice, and music tools all run on the same deal: you describe the outcome, the machine predicts its way there.
Your next step: the film school
Every AI film starts exactly this way — with a clear description of what you want to see. The film school courses turn this skill into finished films.
Two quick questions. Answers check instantly — no grade, just practice.
1. What do image, video, voice, and music AI tools have in common?
The four ingredients still apply — what you want, the background, the limits, the shape of the answer — just pointed at pictures or sound instead of words.
2. Every AI film starts with…
That's the doorway to the film school: a clear description of what you want to see. Everything else is technique.
Describe an image
This proves the skill transfers — no new learning required. Write a four-part prompt for an image: the outcome you want to see, the context and style, the constraints to respect, the format you want back. Run it in any free image tool, or just write it here. Nothing here is required — this one is play.
Your AI starter kit
Make your personal cheat sheet: your best prompts, your tool picks, and your safety rule — one page you'll actually use.
- How to gather your best prompts and rules onto one page you can actually use
- How to pick one real task to try with AI next week
🎙 Theo explains · Read transcript
Time to assemble your AI starter kit: a one-page document with your three best tested prompts, which tool you reach for and when, and your verification habit written down as a rule. This is your capstone — the thing you keep after the course ends. You arrived knowing AI was important and feeling behind. You are leaving with a working method, tested on real tools, that no interface redesign can take away from you.
- Three tested prompts — your best versions from lesson 3 and the practice assignment below.
- Which tool for what — your lesson 2 notes: which tool you reach for, and when.
- Your verification rule — written down from lesson 4, in your own words.
- One thing to try next week — one real task from your life to hand to AI.
Two quick questions. Answers check instantly — no grade, just practice.
1. What is your starter kit, exactly?
It's your personal cheat sheet, written in your own words across this course. Keep it open next to your AI tool and use it.
2. What's the smartest next step after this course?
One real task beats ten more lessons. That's how the skill sticks.
Assemble your starter kit
This is the page you'll keep open next to your AI tool. Everything you saved in the exercises is already below, in your own words. Fill in the last line, copy it, and use it in real life — starting next week.
📋 My AI starter kit
Assembled from your exercises in this course — your words, on one page, ready to use.
Ready to explain AI?
Choose one answer for each question. 10 questions · 8 to pass. After you submit, you'll get a score, explanations, and clear options for what to do next.
1. What best describes what AI does?
AI predicts likely text — it doesn't look up answers or understand like a person does.
2. ChatGPT, Claude, Gemini and Copilot differ mostly in…?
Same underlying idea; they differ in packaging, personality, and strengths.
3. Which set holds the four parts of a strong prompt?
Outcome, context, constraints, format — the four-part formula from lesson 3.
4. The first result misses the mark. What now?
Iteration is the skill: say what to keep, what to change, and what good looks like.
5. "Hallucination" means the AI…?
Hallucination means confidently invented facts — which is why verification matters.
6. Before acting on an AI answer about money or health, you should…?
Confidence is not correctness — verify from a real source before acting.
7. To start comparing the tools in lesson 2, you need…?
Every tool in this course has a free tier, and that's all you need here.
8. The prompting skill also works for…?
The same four-part prompting skill works for image, video, and voice tools too.
9. Your AI starter kit contains…?
Three tested prompts, your tool picks, and a verification rule — one page you keep.
10. Who is responsible for checking AI output before you act on it?
AI reduces effort, but you stay responsible for what you act on.
Try one complete prompt
This is not another quiz question. It's a practical next step: copy the prompt below, paste it into one of the tools from lesson 2, and see what comes back. Zero friction — it's already filled in. Make it yours afterward.
Built from the four-part prompt formula in lesson 3 ↑