A narrated, hands-on workshop · ~60 min · for absolute beginners

AI in One Hour

Press play and I'll walk you through what AI actually is — no jargon, no hype, no homework you can't do. By the end you'll know what's really happening inside these things, what to trust, what to check, and how to ask for what you want. Completely free.

13 chapters · player controls live at the bottom · 100% free

Training text — one small paragraph

The toy writes, one predicted word at a time

Train the toy, pick a starting word, then step.
Step 1: hit Train.
Honest label: this toy learned from the one paragraph on the left, by counting which word follows which. A real language model learns from a vast slice of the internet, with billions of adjustable numbers inside, predicting pieces of words. It is unimaginably bigger — but the core move you're watching is the same one: score every possible next word, then pick.

Ch. 1 / The prediction machine ▶ narrated

It's not a mind. It's a next-word engine.

One sentence explains almost everything you'll ever see an AI do — the brilliance and the failures both: a language model is a machine trained to predict the next piece of text. That's the whole job. Everything else is a consequence.

🎯

It reads, scores, picks

Given everything written so far, it scores every possible next piece of text and picks a likely one. Then it does that again. And again. One piece at a time, until the answer exists.

// the only move it has
📚

It never looks anything up

There is no database of facts inside, no live search by default. Every answer is rebuilt on the spot from patterns it absorbed during training — which is why it can be brilliant and wrong in the same sentence.

// patterns, not lookups
🗣️

Fluency is the skill

What it was optimized for is sounding right — producing text shaped like good text. Truth often comes along for the ride, because true things are common in its training. But truth is the passenger, not the driver.

// remember this one
Try it against the toy: scroll up and auto-write a few sentences. Notice the toy never "decides what to say" — it just keeps picking a likely next word. A real model is doing a vastly richer version of exactly that.

Ch. 2 / Tokens & context ▶ narrated

It reads in chunks, and its desk has edges.

Two small mechanics explain a lot of everyday AI behavior. First: models don't see words or letters — they see tokens, little chunks of text. Second: they can only "see" a limited amount at once — the context window. Type below and watch the chunking.

Honest label: this chopper is a teaching approximation. Every real model chops text with its own learned rules, so the exact pieces differ — but the idea holds: common words survive whole, rare words get split into pieces.
The context window — everything the model can see right now
oldest messages…
the conversation so far
your latest message
its reply

When a chat runs long, the oldest part slides off the left edge — and the model genuinely no longer sees it. Sizes vary by product and change often, so check yours; the shape of the problem is always this bar.

Ch. 3 / Your first skill ▶ narrated

The clear ask: situation, task, shape.

Here's the skill that changes your results today: stop asking like a search box, start briefing like you'd brief a capable helper. Three parts — the situation, the task, and the shape you want back.

The weak ask — what most people type

write an email asking for a day off

The clear ask — same request, briefed

SITUATION: I'm a junior member of a small, friendly team. My manager is busy and reads on her phone. I need this Friday off for a family matter I'd rather not detail. TASK: Write the email requesting Friday off, offering to move my Thursday tasks. SHAPE: Under 90 words. Warm but professional. Subject line included.
Do this now: open any AI assistant in another tab, pick one small real thing from your week, and send it as a three-part ask — situation, task, shape. Then send the lazy one-liner version and compare. You'll feel the difference immediately, and that feeling is the skill installing itself.

Ch. 4 / Hallucination ▶ narrated

Why it makes things up — with a straight face.

All six statements below are written in the same calm, confident, "helpful assistant" voice. Three of them are false. Click the ones you think are false, then check. The point isn't trivia — it's that the voice gives you nothing.

The lesson: a language model produces the confident-textbook tone because that's the pattern good answers have — whether or not the content is true. You cannot detect truth from tone. Ever.

Ch. 5 / Home turf ▶ narrated

What AI is genuinely great at.

The pattern behind every one of these: language goes in, language comes out, and you can judge the result the moment you see it. No hidden facts to trust — just work you'd recognize as good or bad on sight.

📝

Drafting

The blank page killer. A rough first version of anything — email, post, plan, speech — in seconds, for you to shape.

// first draft, not last word
🔁

Rewriting

Same content, different clothes: friendlier, shorter, more formal, simpler. Tone transforms are its native language.

// language in, language out
🗜️

Summarizing

Four pages of notes into five bullets. Long email thread into "what do they actually want?" You verify against the source you already have.

// you hold the original
🧑‍🏫

Explaining

An endlessly patient teacher that will re-explain anything, at any level, with any analogy, and never sigh at you.

// ask again, dumber — it's fine
💡

Brainstorming

Twenty gift ideas, ten headline options, five ways to phrase bad news. Ideas are cheap now; your taste picks the winner.

// volume, then your judgment
🧱

Structuring

Turning your messy brain-dump into an outline, a table, a checklist, a plan with steps. Order from chaos, instantly.

// shape-shifting your own words

Ch. 6 / The other side ▶ narrated

What it's bad at — sort these and see.

Six everyday asks. For each one: is this AI's home turf, or a risky ask it will answer confidently anyway? Sort them, then check.

The tell: the risky asks all need a live, exact, checkable fact — a current date, a real phone number, a precise count. The safe asks are all language work on material you can judge yourself.

Ch. 7 / The habit ▶ narrated

One verification habit you'll actually keep.

Before using any AI answer, ask two questions: does it matter if this is wrong? and can I check it in two minutes? That gives every answer a lane. Practice on these six:

🟢 Enjoy — low stakes, wrongness is harmless or visible. 🟡 Check first — a real fact you'll rely on: confirm it at the source before it leaves your hands. 🔴 Human expert — health, law, money: AI can help you prepare questions, but a qualified person makes the call.

Ch. 8 / Build a better ask ▶ narrated

Five ingredients. Flip them off and feel why.

Here's a real ask — a note to a landlord about a leaking tap — built from five ingredients. All five start on. Turn them off one at a time and watch the ask degrade, with an honest warning about what you just lost.

Ch. 9 / Do it yourself ▶ narrated

The fifteen-minute loop that makes it stick.

Listening builds understanding. Doing builds skill. Here's the whole exercise — do it once this week and you'll be ahead of most people who "use AI".

STEP 1

Pick three real tasks

From your actual week: one 🟢 (a draft or rewrite), one 🟡 (something with a fact you'd rely on), one that's 🔴-adjacent (where you'd want an expert).

STEP 2

Write clear asks

Situation, task, shape — for each one. Add the honesty line: "if you're missing a detail, ask me instead of inventing it."

STEP 3

Judge and verify

Judge the 🟢 on sight. Check the 🟡's fact at its source — actually do it, and notice whether the AI was right. For the 🔴, notice where the machine's help should stop.

STEP 4

Keep a "works for me" note

One page: asks that worked, asks that flopped, facts it got wrong. A month of this and you'll have something no course can sell you — your own evidence.

Ch. 10 / Memory & privacy ▶ narrated

The blank slate, and the postcard rule.

Two things beginners assume that aren't true: that the AI remembers them, and that the chat is a private diary. Here's the honest picture.

🧼

New chat = blank slate

By default, a fresh conversation knows nothing about you or your last chat. Some products add optional memory features — that's a setting you control, not something the model does on its own.

// drifting chat? just start fresh
🧊

It isn't learning you live

The model's internal numbers do not change while you chat. Training happens separately, in big offline batches, before you arrive. Your conversation steers the answer — it doesn't reprogram the machine.

// context, not brain surgery
📮

The postcard rule

Treat anything you type as a postcard, not a sealed letter: never paste passwords, ID numbers, other people's private details, or your employer's secrets. Check your provider's data policy — they differ.

// the one privacy rule to keep
Why this is good news: the blank slate cuts both ways. A conversation that's gone weird or stuck in a rut carries no grudge — open a new chat and you get the model at its best again, every time.

Ch. 11 / The real world ▶ narrated

A normal week, with a copilot in it.

This is what "using AI" actually looks like when the novelty wears off — small, boring, genuinely useful moments, each one running through the lane check from chapter seven.

✉️

Draft it

The email you've been avoiding — first version in 20 seconds 🟢

🗜️

Shrink it

Meeting notes → five bullets you'll actually read 🟢

🧑‍🏫

Decode it

A scary form or letter, explained in plain words 🟡 verify the specifics

You decide

Anything with real stakes still crosses your desk — and stays there

Ch. 12 / Recap & check ▶ narrated

One hour in. Let's prove it stuck.

Five questions. If you listened, you know these cold.

0/5

Where this course sits

This was Level 0 of the AI Ladder.

Everything in this hour is complete on its own — the mental model, the habit, the clear ask are yours now, free, no catch. If you want to keep climbing, this course is the ground floor of a series built in exactly the format you just experienced: narrated, hands-on, honest about what's simulated and what's real.

🪜

Level 0 — this course

What AI is, what to trust, the verification habit, the clear ask. Free, because everyone should have this floor under them.

// you are here
🧗

The next rungs

The paid courses go deeper on the craft: getting reliable work out of AI day after day, and wiring it into your actual documents, work, and tools — same narrator, same style, more reps.

// same format, more depth
🧭

No pressure, honestly

The current lineup and prices live on the store shelf where you found this course. If this free hour was all you needed — that's a win, and we mean that.

// the habit was the point

Transcript

Now playing
Ready — press play to begin
0.95×
🎙