Jump to →

Hero Page Intro: Welcome / chapter opener. Big type, soft glow, tiny progress meter.

P01 · Stage 1 · Explorer

Welcome to The PLAIBOOK

You used ChatGPT a few times and have no idea what just happened. This is the playbook.

2% complete

Myth vs Reality: Two-column tear-down. Red wash on left, cyan wash on right.

What you think

  • × It thinks
  • × It knows things
  • × It remembers you
  • × It's always right

What it is

  • It predicts text
  • It pattern-matches
  • It forgets everything
  • It hallucinates

Autocomplete Steps: 4-step vertical visual explainer with growing scale.

Same mechanism. Bigger scope.

01Phone keyboard
predicts → next word

I'm running late

02Small model
predicts → next sentence

The dog chased the ball across the yard.

03Big model
predicts → next paragraph

Dear Sarah, thanks for sending over the proposal. I reviewed it last night and have a few thoughts I'd like to share before our meeting.

04Frontier model
predicts → entire conversation

Help me launch a business → [full back-and-forth, plans, edits, follow-ups, all from the same prediction loop]

Every step is the same trick: guess what comes next. The only thing that grows is how much it can guess at once.

Temperature Dial: Slider users actually drag. Output blurb changes from safe → wild.

What it is

Temperature is a setting that decides how creative or boring an AI's answer will be.

Why this even exists

When AI writes, it doesn't pick whole sentences. It picks one word at a time. For every next word, it has a list of guesses ranked by how likely they are. The safest guess is always at the top.

But what if you don't want safe? What if you're writing a poem, brainstorming product names, or want a fresh take? Temperature lets you tell the AI: “don't just take the obvious word — surprise me.”

Watch it work — we're freezing time on ONE word

Heads up: the AI is going to do this again for every single word it writes. We're zooming in on just one moment so you can actually see what it's doing under the hood. In real life, this whole process repeats hundreds of times to produce one paragraph — way too fast to watch.

Imagine the AI has written this much so far. Right now it's picking the very next word →

“The cat sat on the mat …”

(After this word is locked in, the AI will run the same picker again for the next word, then the next, then the next — until it's done.)

Temperature

ICE COLD · 0.00

0.0 safest0.7 default1.5 wildest

Here are the words the AI was considering, ranked by likelihood. The dial decides how far down the list it's allowed to look:

mat
62%← picked
floor
18%
rug
10%
couch
6%
throne
3%
spaceship
1%

At 0.0 (Ice cold): AI is only allowed to pick the #1 most likely word. “Mat” wins every time. Boring, but predictable — and predictable is what you want for facts and code.

When to use which setting

Use LOW (0.0–0.3) for

  • · Math, code, factual answers
  • · Customer service replies
  • · Anything where being “wrong but creative” is bad

Use DEFAULT (0.5–0.8) for

  • · Most everyday writing
  • · Emails, summaries, explanations
  • · When you don't know what to pick — start here

Use HIGH (0.9–1.2) for

  • · Brainstorming names, ideas, headlines
  • · Creative writing, poetry, marketing copy
  • · When you want fresh angles instead of obvious ones

Use WILD (1.3+) for

  • · Experiments only
  • · When you want chaos to spark an idea
  • · Expect broken sentences sometimes

Remember this

Most chat tools (ChatGPT, Claude) hide the temperature dial — but it's set around 0.7 by default. When you start using the API or developer tools, this slider is one of the first things you'll see. Now you know what it does.

Hallucination Quiz: Three answers, one is fake. Click reveals correct + XP toast.

Which one is hallucinated?

Goldfish Diagram: Animated bubbles drifting away. Pure metaphor visual.

🐟

Each chat starts from zero

Platform Persistence Grid: Sortable feature matrix with check / x marks.

FeatureChatGPTClaudeGemini
Custom Instructions
Memory across chats×
Project workspaces××
File persistence×

Context Window Desk: Draggable papers metaphor. Shows what falls off the edges.

Your AI's desk

← falls off the edge

Prompt Levels: Five tiers stacked like floors of a building.

L1Ask Clearly
L2Add Context
L3Roles & Examples
L4Chain Prompts
L5Build Templates

Before / After Prompt: Bad prompt struck through. Good prompt with output preview.

Before

Help me with marketing.

→ generic, unusable output

After

I run a dog grooming shop in Austin. Give me 5 Instagram post ideas to attract new customers this summer. Casual tone, under 100 chars each.

→ specific, on-brand, immediately useful

Tool Showdown: Honest grid: ChatGPT vs Claude vs Gemini, traffic-light cells.

ChatGPT

$20/mo

Context: 128k

Best at: Ecosystem

Claude

$20/mo

Context: 200k

Best at: Writing & code

Gemini

$20/mo

Context: 1M

Best at: Long docs

Pricing Tier: ROI breakdown with hourly-rate calculator vibe.

ROI Math

$20/mo pays for itself in...

1 hr

$25/hr

48 min

$75/hr

16 min

$200/hr

6 min

Stage Progress: Horizontal rank tracker — Explorer → Earner with XP.

XP 612 / 2,200Rank: Navigator
ExplorerDriverNavigatorBuilderConnectorEarner

Mastery Bar: One bar fills as you finish pages. Tiers unlock as you cross each marker. Click to add a point.

Current tier

Driver

Mastery

28/100

Explorer

Driver

Navigator

Builder

Connector

Earner

6 steps until Navigator

Commitment Card: Declaration card with signature line. Final-page energy.

P98 · Capstone

I am building...

_____________________________________
_____________________________________
_____________________________________

Signed

Flashcard Flip: Click to flip. Glossary mode for the 50+ AI terms.

Drag & Drop Sorter: Sort items into 'Safe to share' vs 'Never share' columns.

Items

Your home address
A blog post draft
Client SSN
Marketing copy

Safe to share

Never share

Stepper Wizard: Multi-step prompt builder. Forward / back, validation, progress dots.

1
2
3
4

Step 1

Pick a tool

Code Type-Out: Self-typing code block. Feels like AI is writing it live.

Token Counter: Live token counter as you type. Cost ticks up in real time.

Chars

55

Tokens

14

Cost

$0.000042

Vector Search Demo: Type a query → see keyword vs semantic results split-screen.

Keyword search

How to make a paper airplane

Make money online: 5 ideas

Vector search

Best AI side hustles 2025

Pricing your freelance services

Passive income with prompts

Decision Tree: Interactive branching choices. Click yes/no through to a recommendation.

Are users mostly viewing or inputting?

Animated Stat Counter: Numbers that count up on scroll-into-view.

0

Builders

0

Prompts run

0

Total XP

Streak Calendar: 30-day plan tracker. Heatmap-style daily check-ins.

11-day streak

30-day plan

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

Radar Chart: Pure SVG radar showing your skill across the 6 stages.

ExploreDriveNavigateBuildConnectEarn

Accordion FAQ: Expand/collapse cards for the troubleshooting page.

Your prompt is too vague. Add context, constraints, examples.

Toast Notification: XP earned, badge unlocked, level up — slide-in toast.

Tabbed Walkthrough: Platform-specific instructions. Tab between ChatGPT / Claude / Gemini.

Settings → Personalization → Custom Instructions. Two boxes: about you, response style.

Reveal Cards: Click each card to flip from 'myth' to 'reality'. Track count.

Revealed 0 / 4

Comparison Slider: Drag the divider — left side AI image, right side human design.

AI Generated
Human Made

Annotated Screenshot: Numbered hotspots on a UI screenshot. Hover for callouts.

Confetti Complete: Page-completion celebration. Hit the button.

Prompt Builder: Fill the variables, watch the prompt assemble live below.

Assembled prompt

You are a senior copywriter. Write Instagram captions for Gen Z skincare buyers. Keep the tone playful.

Leaderboard: Community leaderboard mock. Your row highlighted.

#1vaibe_kelsey
2,180 XP
#2promptdad
1,940 XP
#3you
1,612 XP
#4tokenwitch
1,505 XP
#5agentanon
1,402 XP

END · concept sheet · 15 / 15