Summer 2026

AI 101

A free 8-week curriculum to build real AI fluency — no coding required. Focused on business, entrepreneurship, and building a portfolio of real projects.

8 Weeks
~5 hrs / week
No coding needed
6 portfolio projects

How This Works

A practical, hands-on guide to building real AI fluency

📐 Structure

Each week has a 1-hour lesson to work through, followed by 3–4 hours of hands-on exercises and a mini-project to complete independently. Weekends are for the project.

🧰 Tools

Every tool used in this curriculum is either free or has a free tier sufficient for the exercises. No paid subscriptions required. No coding required.

🏗️ Projects

Each week produces a tangible deliverable — a market analysis, a business plan, a pitch deck, a website. By Week 8, you'll have a portfolio of 6+ real projects.

🧭 Philosophy

AI is a thinking partner, not a replacement for thinking. Every exercise emphasizes judgment, verification, and adding human insight on top of AI output. The goal is fluency, not dependency.

1

What Is AI, Really?

Foundations — demystify the technology and start using it
  • Understand what AI is (and isn't) — separate hype from reality
  • Learn the difference between generative AI, machine learning, and traditional software
  • Get comfortable with the basic interaction model: prompt → response → refine
  • Understand that AI "hallucinates" — and why verification always matters
ChatGPT Google Gemini Claude DeepSeek
  • The Comparison Test: Ask the same question to ChatGPT, Gemini, Claude, and DeepSeek. Compare answers side by side. Which was most helpful? Most accurate? Notice any differences between the US-made models and DeepSeek (China)? Why might they differ?
  • The Hallucination Hunt: Ask each AI about a real but obscure topic you already know about (a local business, a family member's profession, a niche hobby). Fact-check the responses. Find the mistakes.
  • Prompt Iteration: Start with a vague prompt ("Tell me about electric cars") and refine it three times, getting more specific each round. Document how the output quality improves with better prompts.
  • The Explanation Ladder: Ask an AI to explain blockchain at five levels: to a 5-year-old, a 10-year-old, a high schooler, a college student, and an expert. Notice how it adapts.
  • When an AI gives you a wrong answer, whose fault is it — the AI's, or the person who believed it without checking?
  • What jobs do you think AI will change the most in the next 10 years? What jobs won't change much?
  • If you could have an AI assistant that did one thing perfectly, what would you want it to do?

Project: "AI Myth vs. Reality" One-Pager

Write a one-page document (using an AI to help draft it, but with your own thinking and edits) that busts 5 common AI myths for your classmates. Examples: "AI is smarter than humans," "AI will take all jobs," "AI always tells the truth." For each myth, explain the reality. This becomes Portfolio Piece #1.

⏱ ~5 hours total this week
Detailed Lesson Plan — Week 1

If you're new to AI and want a slower, hand-held walkthrough, read this section before doing the exercises above. Total reading time: about 15 minutes.

Mental Model: What's Actually Happening

A modern AI chatbot is a "next-word predictor" trained on a huge chunk of the internet. It doesn't know things the way you do — it has learned patterns of language. When it sounds confident and wrong, that's not lying; it's predicting plausible-sounding text. This is why we say it "hallucinates."

Think of it less like HAL 9000 and more like a charming friend who has skimmed every book in the library at 3 a.m. and will absolutely make stuff up rather than admit they don't remember chapter 4.
Step 1: Set Up Your Accounts (~20 minutes)
  1. Go to chat.openai.com and create a free account. The free tier is plenty.
  2. Go to claude.ai and sign up. Claude tends to be the best free option for longer writing tasks.
  3. Go to gemini.google.com — sign in with any Google account.
  4. Go to chat.deepseek.com and sign up. This is a Chinese-made model — useful for the comparison exercise.

Privacy tip: By default, free chatbots may use your conversations to train future models. In each tool's settings, look for "Data Controls" or "Improve the model for everyone" and turn it off if you'd rather not contribute. Don't paste anything truly sensitive (medical info, passwords, your social security number) into any AI ever.

Step 2: Your First "Real" Conversation (~15 minutes)
  1. Open ChatGPT. Type: "Explain what you are and how you work, in plain English, like I'm a smart 8th grader."
  2. Read the answer. Then ask: "What are three things you're bad at?" — notice how it actually tells you its weaknesses.
  3. Now ask the exact same two questions in Claude and Gemini. Compare the personalities. They are noticeably different.
Step 3: Run the Hallucination Hunt

Pick a topic only you would know well. Examples: a small local restaurant in your town, your grandfather's career, the rules of a niche video game, a school club you ran. Ask each AI for "5 facts" about it.

  • Some "facts" will be right. Some will be confidently wrong. Write down both.
  • This is the single most important lesson of the entire 8 weeks: verification is your job, not the AI's.
"It is a tale told by an AI, full of citations and confidence, signifying nothing." — Shakespeare, probably (verify before quoting).
If You Get Stuck
  • "My AI keeps refusing to answer." Try rephrasing — say what you actually need it for ("I'm doing a school project on…"). Or try a different model; they have different policies.
  • "The answers all look the same." Push harder. Ask follow-up questions. The first answer is usually generic; the third or fourth is where the gold lives.
Going Deeper (Optional)
2

The Art of Prompting

Learn to communicate with AI like a pro
  • Master the core prompting techniques: context, role, format, constraints, examples
  • Learn to break complex tasks into step-by-step instructions
  • Understand "few-shot" prompting — teaching AI by example
  • Discover that prompt engineering is the single highest-leverage AI skill
ChatGPT Claude
  • Role Play Prompting: Ask the AI to respond as different experts — a marketing executive, a scientist, a comedian, a strict editor. Notice how the role changes the output style and quality.
  • Format Control: Take one topic and ask for the response as: a bulleted list, a formal email, a tweet thread, a table, a pros/cons analysis. Learn that format is a lever you control.
  • Chain-of-Thought: Ask the AI to "think step by step" when solving a logic problem or making a recommendation. Compare the answer quality to when you don't ask for reasoning.
  • The Prompt Battle: Try getting the AI to produce the best possible product description for the same imaginary product as a friend or study partner. Compare prompts and outputs. Discuss what made the winning prompt better.
  • Why is "garbage in, garbage out" even more true with AI than with Google searches?
  • A friend says "I just type whatever and it works fine." What is that friend missing out on?

Project: Personal Prompt Library

Build a Google Doc or Notion page with your "Top 10 Prompts" — reusable prompt templates for tasks you care about (homework help, brainstorming ideas, summarizing articles, writing emails, etc.). Each prompt should include the template, an example, and a note on why it works. This is a living document you'll keep expanding all summer.

⏱ ~5 hours total this week
Detailed Lesson Plan — Week 2

Last week you learned that AI is a powerful but unreliable friend. This week you learn how to actually talk to that friend. Prompting is the difference between getting a generic answer and getting something genuinely useful.

The 5-Part Prompt Recipe

Almost every great prompt has some combination of these five ingredients. Memorize them. They will serve you for the rest of the curriculum.

  1. Role — who is the AI being? "You are a senior magazine editor."
  2. Context — what's the situation? "I'm writing a 500-word op-ed for a high school newspaper."
  3. Task — what do you actually want? "Critique my draft below and suggest three rewrites for the opening line."
  4. Format — how should the answer look? "Give your critique as a bulleted list, then the three openings as numbered options."
  5. Constraints / Examples — what to do or avoid. "Keep each opening under 20 words. Avoid clichés like 'In today's world.'"
A vague prompt to a powerful AI is like asking Gandalf for "directions." Be the hobbit who brings a map.
Step-by-Step: Build Your First "Power Prompt"
  1. Pick a real task you have this week (an essay, an email, a planning question).
  2. Write the lazy version first: a one-line ask. Send it to Claude. Save the response.
  3. Now rewrite using all 5 ingredients above. Send the new prompt to a fresh Claude chat (so it doesn't remember the old one).
  4. Compare. The difference is usually night and day. That is your prompting skill leveling up in real time.
Few-Shot Prompting (Teaching by Example)

Instead of describing what you want, just show it. Pattern:

  • "Here are two examples of the kind of headline I want:"
  • "Example 1: …"
  • "Example 2: …"
  • "Now write 5 more in the same style for this topic: ___"

This single trick will outperform almost any other technique. The model is a pattern-matcher; give it a pattern.

Chain-of-Thought (Make It Show Its Work)

For anything that needs reasoning — comparisons, math, decisions, recommendations — add this magic phrase: "Think step by step before giving your final answer." Output quality often jumps noticeably. (Newer "reasoning" models do this automatically, but the phrase still helps with anything else.)

Common Mistakes
  • Treating it like Google. Google rewards keywords. AI rewards full sentences and context.
  • Asking for too much in one shot. Break big tasks into 2–3 prompts. Plan first, then draft, then edit.
  • Not telling it your audience. "Explain X" vs. "Explain X to my grandma who used to be a nurse" produces wildly different outputs.
  • Accepting the first answer. Always say "Now do that again, but make it sharper / shorter / more specific / less generic."
"Do or do not, there is no try." — Yoda. With AI, it's: do, then iterate, then iterate again.
Going Deeper (Optional)
3

AI-Powered Research & Analysis

Turn AI into the world's best research assistant
  • Learn to use AI for rapid, credible research — with source verification
  • Understand how Perplexity AI and NotebookLM differ from ChatGPT for research
  • Practice synthesizing information from multiple sources into original analysis
  • Build the habit of always asking: "Is this actually true?"
Perplexity AI Google NotebookLM ChatGPT (Search)
  • Source Showdown: Research "the future of electric vehicles" using Perplexity AI (which cites sources) vs. ChatGPT. Which gives you more trustworthy, verifiable information? Click through the citations — are they real?
  • NotebookLM Deep Dive: Upload 3–4 articles about a trending business topic into NotebookLM. Use it to generate summaries, find connections between articles, and create an "audio overview" podcast-style summary.
  • Competitive Analysis: Pick two real competing companies (e.g., Nike vs. Adidas, Spotify vs. Apple Music). Use AI to research each company's strengths, weaknesses, market share, and recent moves. Organize into a comparison table.
  • Fact-Check Relay: Have the AI make a set of 10 specific claims about a topic. Then verify each claim using Perplexity or Google. Score the AI's accuracy.
  • If AI can do research in 30 seconds that used to take 3 hours, what becomes the new valuable skill?
  • How can you tell the difference between a good source and a bad one? Does AI help or hurt with that?

Project: Market Opportunity Analysis

Pick an industry or trend you're curious about (e.g., AI tutoring, sustainable fashion, esports, plant-based food). Use Perplexity and ChatGPT to research market size, key players, growth trends, and consumer demographics. Compile a 2-page "Market Opportunity Brief" — the kind a real startup founder would write before launching a product. Portfolio Piece #2.

⏱ ~5 hours total this week
Detailed Lesson Plan — Week 3

This week is where AI starts to feel like a superpower. You can do in 30 minutes what used to take a research analyst a full afternoon. The catch: an unverified shortcut is worse than no shortcut at all.

Why Different Tools for Different Jobs
  • ChatGPT / Claude — great at synthesizing what they already "know," but their knowledge has a cutoff date and they can hallucinate sources.
  • Perplexity — built for research. Every claim links to a real, clickable source on the web. Use it whenever you need facts you can defend.
  • NotebookLM — you give it the sources (PDFs, articles, YouTube links). It only answers from those documents. Almost zero hallucinations because it's grounded in your material.
ChatGPT is the friend who tells great stories. Perplexity is the friend who shows up with footnotes. NotebookLM is the friend who actually read the book. Use the right friend for the job.
Step-by-Step: Perplexity Walkthrough (~15 min)
  1. Go to perplexity.ai. No sign-up needed for basic use.
  2. Ask a research question with a date or specifics: "What were the three biggest moves in the EV market in the last 6 months, and why do they matter?"
  3. Read the answer. Now click every numbered citation. Open each source. Does the source actually say what Perplexity claims it says?
  4. Use the "Related" follow-up suggestions — they're often better than what you would've thought to ask next.
  5. Try the "Focus" toggle: switch from "Web" to "Academic" and re-ask. Notice how the source quality changes dramatically.
Step-by-Step: NotebookLM Walkthrough (~20 min)
  1. Go to notebooklm.google.com and sign in with a Google account.
  2. Click "New notebook." Upload 3–4 sources: PDF reports, article URLs, YouTube video links, even your own notes.
  3. In the chat, ask: "What are the three biggest themes that show up across all these sources?"
  4. Try the "Audio Overview" feature — it generates a surprisingly good ~10 minute podcast where two AI hosts discuss your sources. Listen on a walk.
  5. Every answer cites the exact source and passage. Click it. Verify it. This is research with training wheels you should never take off.
Source Quality Quick Filter

When AI cites a source, ask yourself the CRAAP test (real acronym, sorry):

  • Currency — is it recent enough for the question?
  • Relevance — does it actually address what you asked?
  • Authority — who wrote it? A research institute, a company blog, or a random Reddit post?
  • Accuracy — can you cross-check the numbers somewhere else?
  • Purpose — is the source trying to inform you, sell you something, or persuade you politically?
The Sherlock Move

When you have a finding from one AI, paste it into a different AI and ask: "Find the three weakest claims in this analysis and explain what evidence would be needed to verify them." Two AIs disagreeing is how you find the soft spots in a story.

"It is a capital mistake to theorize before one has data." — Sherlock Holmes. The 21st century version: it is a capital mistake to ship a market analysis before you've clicked a single source.
Going Deeper (Optional)
4

AI for Writing & Communication

Write faster, write better — without losing your voice
  • Use AI as a writing partner: brainstorm, draft, edit, refine
  • Learn the critical difference between "AI writes for you" vs. "AI helps you write better"
  • Practice using AI for different writing formats: emails, essays, social media, business copy
  • Develop an editing workflow: draft → AI feedback → human revision
Claude Grammarly ChatGPT Hemingway Editor
  • The Rewrite Ladder: Write a short paragraph about something you care about. Then ask AI to make it more concise, more persuasive, more formal, and more casual. Study what changed each time.
  • AI as Editor: Write a one-page essay on your own (no AI). Then paste it into Claude and ask: "What are the three weakest parts of this essay and how can I improve them?" Revise based on the feedback.
  • Tone Matching: Find a blog post or article whose writing style you admire. Paste a sample and ask the AI to analyze the tone, then help you write something in a similar style on a different topic.
  • Cold Email Challenge: Draft a cold email to a (hypothetical) local business owner proposing a summer internship. Use AI to iterate on the email until it's genuinely compelling. Would you actually send this?
  • If everyone starts using AI to write, what makes one person's writing stand out from another's?
  • When is it OK to use AI for writing? When is it not? Where is the line between a tool and a crutch?

Project: Business Blog Post

Write a 600–800 word blog post about the market opportunity you researched last week. The catch: write the first draft entirely yourself. Then use AI to get feedback, improve it, and polish it. The final product should be clearly yours, but demonstrably better because of how you used AI in the process. Document your drafting process (before/after). Portfolio Piece #3.

⏱ ~5 hours total this week
Detailed Lesson Plan — Week 4

The most important sentence of this entire week: let AI improve your writing, not replace it. Anyone can paste a topic into ChatGPT and get 600 words of perfectly bland mush. The skill — and the value — is in keeping your voice while making everything around it sharper.

The Editing Workflow (Use This Forever)
  1. Brainstorm with AI. Dump messy ideas. Ask for outlines, angles, counter-arguments. Don't let it write yet.
  2. Draft yourself. Bad first drafts are fine. Voice lives in the rough edges.
  3. AI as critic, not author. Paste your draft and ask: "What are the three weakest paragraphs and why? Don't rewrite — just diagnose."
  4. Revise yourself based on the diagnosis.
  5. Final polish pass. Now you can ask AI for line edits, tighter phrasing, grammar fixes.
If your essay sounds like everyone else's essay, you've used AI as a ghostwriter. If it sounds like you, only sharper, you've used AI as an editor. The difference is everything.
Step-by-Step: Hemingway Editor (~10 min)
  1. Go to hemingwayapp.com. No account needed for basic use.
  2. Paste your draft. It color-codes problems: yellow = hard sentence, red = very hard, purple = simpler word available, green = passive voice, blue = adverbs (use sparingly).
  3. Goal: get to "Grade 8" or lower readability for most writing. Even great essayists write at this level — clarity ≠ dumbing down.
  4. Don't accept every suggestion. Some long sentences are long on purpose. Trust your ear.
Step-by-Step: The "Three-Pass" Claude Edit
  1. Pass 1 — Diagnosis only: "Below is my essay. Don't rewrite it. List the 3 strongest moments and the 3 weakest moments, and explain each in one sentence."
  2. Pass 2 — Targeted improvement: Pick one weak section. Ask: "Suggest 3 different ways I could rewrite this paragraph, each with a different angle. Keep my voice — I tend to use [describe your style: short sentences, dry humor, etc.]."
  3. Pass 3 — Line polish: "Suggest 5 sentence-level tightenings. For each, show the original and the proposed change side by side." Accept what helps; reject what flattens.
Keeping Your Voice

AI defaults to corporate-LinkedIn neutral. If you don't fight it, you'll write like everyone else. Counter-prompts:

  • "Don't sand off the personality. Keep the dry humor and the contractions."
  • "Avoid these phrases: 'in today's fast-paced world,' 'leverage,' 'unlock,' 'delve,' 'tapestry.'"
  • "Match this writing sample for tone: [paste a paragraph you wrote]."
If your essay contains the word "delve" and you didn't put it there, an AI did. It's the duct tape on the chatbot's bumper — a tell.
The Ethics Conversation You Should Have With Yourself

Different contexts have different rules. Sort your writing tasks into three buckets:

  • Yours alone — anything you're being graded or evaluated on as a measure of your ability. AI for brainstorming and editing only; never for drafting.
  • Collaborative — work emails, marketing copy, blog posts. AI is a teammate. Disclose if asked.
  • Functional — meeting notes, summaries, formatting. AI does the heavy lifting; you spot-check.
Going Deeper (Optional)
5

AI for Visual Content & Presentations

Design like a pro with zero design experience
  • Use AI to generate images, logos, and visual concepts
  • Build professional presentations using AI-powered tools
  • Understand the basics of visual communication and slide design
  • Learn about AI image generation ethics: attribution, deepfakes, and copyright
Canva (AI features) Gamma AI ChatGPT (DALL-E) Google Slides + Gemini
  • Image Prompt Engineering: Use DALL-E (via ChatGPT) to generate product mockups or brand imagery. Learn how specific visual descriptions (lighting, style, composition) dramatically change output quality.
  • Rapid Deck Building: Take your Market Opportunity Brief from Week 3 and paste it into Gamma AI. Generate a presentation in under 2 minutes. Then critique it: what's good? What would you change?
  • Canva + AI: Use Canva's Magic Design to create a social media post promoting your hypothetical business. Experiment with different templates, AI-generated copy, and layouts.
  • Before & After: Create a presentation manually in Google Slides (10 minutes), then create the same one with Gamma AI (2 minutes). Compare quality, speed, and what each approach is better for.
  • AI can now generate photorealistic images of people who don't exist. What are the dangers of that?
  • If AI can make anyone look like a great designer, what new skills become the differentiator?

Project: Investor Pitch Deck

Create a 10-slide pitch deck for a business idea (real or invented) using Gamma AI as a starting point, then customize in Canva or Google Slides. Include: problem, solution, market size, competition, business model, team, and a clear ask. This is the kind of deck real founders show to investors. Portfolio Piece #4.

⏱ ~5 hours total this week
Detailed Lesson Plan — Week 5

Five years ago, "I'm not a designer" was a real excuse. This week, that excuse expires. With AI, the bottleneck moves from making visuals to knowing what looks good — which is a much faster skill to build.

The Image-Generation Mental Shift

Most beginners type three words: "logo for bakery." Then they wonder why the result looks like every other AI bakery logo on Earth. A great image prompt has five layers:

  1. Subject — what's literally in the image. "A vintage hand-drawn logo of a rolling pin and wheat sheaf, intertwined."
  2. Style — visual genre. "In the style of mid-century European bakery signage, hand-lettered."
  3. Composition — framing. "Centered, circular badge layout, with the bakery name 'Hearth & Crust' in arched serif type."
  4. Color & Lighting — mood. "Warm cream background, deep maroon and golden ochre ink, slightly aged paper texture."
  5. Negative space / "no" list — what to avoid. "Avoid cartoonish style, avoid drop shadows, avoid modern gradients."
Generic prompt in, generic image out. Vague "make me a logo" is the visual equivalent of "make me a song" — you'll get something, and it will be soulless.
Step-by-Step: Gamma AI Pitch Deck (~30 min)
  1. Go to gamma.app and sign up. The free tier gives you 400 "AI credits" — plenty for a deck.
  2. Click "Create new" → "Generate" → "Presentation."
  3. Paste your Week 3 Market Opportunity Brief as the source. Tell it: "10-slide investor pitch. Tone: confident but not hype-y. Audience: pre-seed VCs."
  4. Pick a theme. (The "Oasis" and "Atlas" themes look more professional than the default.)
  5. Let it generate. It will be 80% there in 60 seconds.
  6. The 20% that matters: rewrite every headline yourself. Replace generic stock images with specific ones. Tighten the "Ask" slide — investors stare at that one for ten seconds and decide.
Step-by-Step: Canva Magic Design (~20 min)
  1. Go to canva.com. Free account.
  2. Click the purple "Magic Studio" or just type into the home search: "Instagram post for an artisan bakery launch".
  3. Pick a template you like, then use "Magic Edit" to swap images, "Magic Write" to rewrite copy, and "Magic Resize" to make TikTok / story / square versions instantly.
  4. Export as PNG or PDF.
Slide Design Rules (That Beat Any AI)
  • One idea per slide. If you wouldn't say it out loud as a sentence, it doesn't belong as a bullet.
  • Headlines, not topics. "Revenue Growth" is a topic. "We tripled revenue while halving spend" is a headline. Use headlines.
  • Six words, max, on a line. If you must read your slide, the audience is reading it instead of listening to you.
  • One image, not five. Whitespace is a feature.
  • Numbers are stories. "60% growth" needs context. "60% growth — fastest in the category" is a story.
The Ethics Slide You Should Sit With

AI image tools can now generate convincing photos of real people doing things they never did. This is not theoretical — it's happening in elections, harassment campaigns, and scams.

  • Don't generate images of real, identifiable people without their consent. Ever.
  • Disclose AI-generated images when context matters (journalism, evidence, marketing claims).
  • Learn to spot deepfakes: weird hands, melted backgrounds, mismatched ear/jewelry symmetry, soft text.
Black Mirror is not supposed to be an instruction manual. Don't make Charlie Brooker's job harder than it already is.
Going Deeper (Optional)
6

AI for Business & Automation

Build workflows that run themselves
  • Discover no-code automation: connecting apps and building workflows without programming
  • Understand how real businesses use AI to save time and money
  • Build a simple automated workflow from scratch
  • Learn to think in systems: inputs → process → outputs
Zapier (free tier) Notion AI Otter.ai ChatGPT
  • Workflow Mapping: Pick a repetitive task (e.g., tracking homework assignments, summarizing news, organizing notes). Map it out on paper: what triggers it? What steps happen? What's the output? Now automate it.
  • Zapier Starter: Build a simple Zap — e.g., "When I star an email in Gmail, save it to a Google Sheet and send me a summary via text." Experience the magic of automation.
  • Notion AI Dashboard: Set up a Notion workspace for a hypothetical small business: client tracker, task list, meeting notes. Use Notion AI to auto-summarize and organize.
  • Meeting Simulator: Record a "mock meeting" conversation (with a friend or colleague), transcribe it with Otter.ai, and use the AI-generated summary and action items. See how much time this saves in the real world.
  • If AI can automate a task, does that mean the task was never "real work" in the first place?
  • What's the difference between automating something and understanding something?

Project: Automated Business Dashboard

Design and build a Notion-based "command center" for your hypothetical business. Include pages for: daily tasks, customer research notes (with AI summaries), a content calendar, and a competitive tracker. Set up at least one Zapier automation that feeds data into it. Portfolio Piece #5.

⏱ ~6 hours total this week
Detailed Lesson Plan — Week 6

This is the week most people have their "wait, I can do this?" moment. Automation used to require a programmer. Now it requires patience and a free Zapier account.

The Mental Model: Trigger → Action

Every automation, no matter how complex, is built from this one shape:

  • Trigger: something happens (a new email, a starred message, a row added to a sheet).
  • Action: the system does something in response (send a message, save a row, summarize text).

You can chain actions. Trigger → Action → Action → Action. That's the whole game. Most "complex" workflows are just a chain of these dominos.

If Tony Stark had a J.A.R.V.I.S., you have Zapier and a Notion page. The arc reactor is sold separately.
Step-by-Step: Your First Real Zap (~30 min)

We'll build: "When I label a Gmail message 'AI-101', summarize it with ChatGPT and add the summary to a Google Sheet."

  1. Go to zapier.com. Sign up free (5 Zaps, 100 tasks/month — plenty).
  2. In Gmail, create a label called AI-101.
  3. In Google Sheets, create a sheet with columns: Date, Subject, From, Summary.
  4. In Zapier, click "Create Zap."
  5. Trigger: Gmail → "New Labeled Email" → choose your AI-101 label. Test it (label any email so Zapier can grab a sample).
  6. Action 1: ChatGPT (or "Formatter" → "AI by Zapier") → "Conversation" → prompt: "Summarize this email in 2 sentences focused on the action items: {{body}}". Map the email body into the prompt.
  7. Action 2: Google Sheets → "Create Spreadsheet Row" → map Date, Subject, From, and the AI summary into the four columns.
  8. Test the whole Zap. Turn it on. Label a real email. Watch the magic.
Step-by-Step: Notion AI Workspace (~25 min)
  1. Go to notion.so, sign up, pick "Personal."
  2. Create a new page called "Business Command Center."
  3. Add four sub-pages: Tasks, Customer Notes, Content Calendar, Competitor Tracker.
  4. On Tasks: type /database and pick "Table — inline." Add columns: Title, Status (select), Due, Priority.
  5. On Customer Notes: paste raw notes from a (real or imaginary) interview. Highlight the text and click the AI button. Try "Summarize," "Action items," "Find insights."
  6. On Content Calendar: type /calendar for a calendar view. Add columns for Channel, Status, AI-drafted copy.
  7. Use Notion AI to draft 5 sample social posts directly inside the calendar. Edit them so they sound like a human wrote them.
Step-by-Step: Otter.ai Meeting Capture (~10 min)
  1. Sign up at otter.ai. Free tier gets you 300 minutes/month.
  2. Record a 5-minute "mock meeting" (with a friend, or by reading a transcript out loud).
  3. Otter delivers a full transcript, an AI summary, and an action-item list, automatically.
  4. Now imagine doing this for every meeting in a real job. That's the time-savings calculation.
Where Automation Goes Wrong
  • Automating mess. If your manual process is a disaster, the automated version will be a faster disaster. Fix the workflow on paper first.
  • Silent failures. Always check your Zap history weekly. Things break (an API changes, a token expires) and the bot won't tell you.
  • Over-automation. Some tasks deserve human attention. Not every email is meant to be summarized.
  • Privacy leaks. Don't pipe sensitive client data through unknown third-party AI endpoints. Read the terms.
A poorly built automation is a Rube Goldberg machine that emails you about its own failures at 3 a.m. Build deliberately.
Going Deeper (Optional)
7

Build Something Real

Capstone project — put it all together
  • Combine every skill from Weeks 1–6 into one real, end-to-end project
  • Experience the full cycle: ideate → research → plan → build → present
  • Make something you'd genuinely be proud to show on a college or job application

Choose one of these, or pitch your own idea:

🛒 Launch a Micro-Business

Use AI to find a product opportunity, write copy, design branding in Canva, build a simple one-page site (Carrd.co or Lovable), and create a marketing plan. Go from idea to "ready to launch" in one week.

📰 Publish an AI-Assisted Newsletter

Research a niche topic, use AI to help write and edit 3 issues, design them in Canva or Substack, and send to friends/family. Build a real subscriber list.

🎯 Solve a Real Problem

Identify a genuine problem in your school or community. Use AI to research it, propose a solution, build a presentation, and pitch it to someone who could actually implement it (a teacher, local business owner, etc.).

📱 Build a No-Code App

Use Lovable or a similar AI app builder to create a functional web app that solves a problem you care about — a study group organizer, a local events finder, a personal finance tracker.

  • Monday–Tuesday: Ideation and research. Use AI to brainstorm, validate, and plan.
  • Wednesday–Thursday: Build. Create all deliverables using the AI tools you've learned.
  • Friday–Weekend: Polish and document. Write up what you did, how you used AI, and what you learned.
⏱ ~8 hours total this week
Detailed Lesson Plan — Week 7

Six weeks of skills, one week to assemble them. This is your "Avengers, assemble" moment — every tool from Weeks 1–6 has a role, and the project is the team-up movie.

Day-by-Day Schedule

Monday — Pick & Validate (~2 hrs)

  1. Pick one of the four capstone options or pitch your own.
  2. Use Claude or ChatGPT to brainstorm: "I want to do [X]. Give me 10 sub-versions of this idea, ranging from boring-but-doable to ambitious. Then tell me which 3 are most realistic for one week, given I'm a beginner."
  3. Use Perplexity to find: existing competitors / examples, audience size, evidence the problem is real.
  4. Write a one-paragraph "project brief" — the thing you'll judge yourself against on Friday.

Tuesday — Plan (~2 hrs)

  1. List every deliverable: site, copy, images, deck, etc.
  2. For each deliverable, write the tool you'll use and the prompt you'll start with.
  3. Build a Notion or Google Doc tracker. Treat this like a real project.

Wednesday — Build, Part 1 (~2.5 hrs)

  • Get the ugly version of everything done. Don't polish yet. Done > perfect on day 3.

Thursday — Build, Part 2 (~2.5 hrs)

  • Polish your ugly version. This is where AI really earns its keep — copy editing, image swapping, layout cleanup.

Friday + Weekend — Document & Present (~2 hrs)

  • Write a "process log": which AI tools you used, what worked, what didn't, what you'd do differently.
  • Record a 2-minute video walkthrough of the project. Future-you (and admissions officers) will love that you have one.
Step-by-Step: Lovable for the No-Code App Track (~45 min)
  1. Go to lovable.dev. Free tier includes a few generations a day — enough for a hackathon-style build.
  2. In the prompt box, describe what you want in plain English: "Build a study group organizer where users can post a class, time, and location, and others can RSVP. Use a clean modern design with dark mode."
  3. Lovable generates a working web app. You can chat with it to edit features: "Add a search bar at the top," "Make the RSVP button green when clicked."
  4. When it's working, click "Publish" — you get a live URL you can share.
  5. Reality check: the first version will have bugs. Iterate. Describe each bug like you'd describe it to a friend. "When I click submit, the page goes blank. Fix it."
Step-by-Step: Carrd One-Page Site (~30 min)
  1. Go to carrd.co. Free for up to 3 sites.
  2. Pick a single-column landing template.
  3. Replace the headline. Use Claude to draft 5 headline options. Pick the one that's specific, not clever.
  4. Add: a short pitch (60 words), one image (Canva or DALL·E), an email signup, and one button.
  5. Buy a custom domain on Namecheap (~$10/yr) if you want to look professional. Optional but high-impact for portfolios.
The "Real Audience" Multiplier

Anything you ship to a real audience — even 5 people — multiplies what you learn by 10×. Force it. Email your newsletter to family. Share your site in a Discord. Pitch your "solve a real problem" presentation to an actual teacher or business owner.

"Build it and they will come" is from a movie about ghosts. In real life, you have to also tell them you built it.
How to Know You're Done
  • A stranger could land on your project and understand it in under 30 seconds.
  • Every claim you make is something you could defend if asked.
  • You can describe, in one sentence, what you'd improve with another week.
Going Deeper (Optional)
8

Portfolio, Reflection & What's Next

Package your work and plan your continued learning
  • Assemble all your projects into a portfolio you can share with colleges and employers
  • Reflect on what you've learned — and articulate it clearly (a skill in itself)
  • Understand the AI landscape: what's changing, what to watch, how to keep learning
  • Set a plan for continued AI learning after the summer
  • Build Your Portfolio Site: Use Carrd.co (free) or a Notion public page to create a simple portfolio showcasing your 6 projects. For each project: title, description, what AI tools you used, and what you learned.
  • Write Your "AI Journey" Essay: Write a 500-word personal essay about what you learned this summer. This is the kind of essay that could become a college application supplement. Use AI to help edit — not write — it.
  • The Presentation: Create a 5-minute presentation of your capstone project. Practice delivering it. This is your "demo day" — present it to family, friends, or record a video.
  • AI Ethics Reflection: Write a short piece on one ethical concern about AI that you now understand more deeply than you did 8 weeks ago. Show that you think critically, not just technically.
  • What's the most surprising thing you learned about AI this summer?
  • How has your daily use of technology changed since Week 1?
  • If you had to teach a friend one AI skill, which would it be and why?

Portfolio Piece #6: The Portfolio Itself

A live portfolio page with all your projects, an "about me" section, and your AI Journey essay. Share the link. This is the thing you'll put on applications — tangible proof that you understand AI, can use it effectively, and think critically about it. That puts you ahead of 99% of applicants your age.

⏱ ~5 hours total this week
Detailed Lesson Plan — Week 8

Week 8 is the credits-scene week — but in a Marvel way, not a "everyone leaves the theater" way. The work this week is what turns the last seven weeks from "things I did" into "evidence of who I am."

Step-by-Step: Building Your Portfolio Site (~90 min)

You have two clean options. Pick whichever feels easier.

Option A — Carrd (simplest, ~45 min):

  1. At carrd.co, choose a multi-section template (look for one with "portfolio" or "showcase" in the name).
  2. Add a hero section: your name, one sentence about what you do, and a one-line description of the curriculum.
  3. Create a section for each project (6 total). Each section needs: title, 1–2 sentence description, the AI tools you used, what you learned, and a link to view the project.
  4. Add an "About Me" section and a contact email or social link.
  5. Publish to a free yourname.carrd.co URL — or upgrade ($19/yr) to use a custom domain.

Option B — Notion public page (~30 min):

  1. Build a Notion page with the same six project sections.
  2. Click "Share" → "Publish to web." Copy the URL.
  3. For a cleaner URL, run it through Potion.so or Super.so (free trials).
A public portfolio is the "I, for one, welcome our new AI overlords" moment — except instead of welcoming them, you're showing you can actually drive them.
Step-by-Step: Writing the "AI Journey" Essay
  1. Don't start with AI. Open a blank doc. Write three things, by hand if possible:
    • The moment in these 8 weeks when something genuinely surprised you.
    • The mistake you made that taught you the most.
    • One belief about AI you held in Week 1 that you no longer hold.
  2. Now draft a 500-word essay around those three answers. Bad first draft is fine.
  3. Now open Claude. Paste the draft. Ask: "Diagnose the three weakest paragraphs. Don't rewrite them — just tell me why they're weak."
  4. Revise yourself. Repeat until the essay sounds like the way you actually talk.
  5. Final pass: "Suggest 5 sentence-level tightenings. Show before/after." Accept the ones that improve clarity; reject the ones that flatten your voice.
The Five-Minute Demo Day Talk

Structure your capstone presentation like this:

  1. (0:00–0:30) The hook. Open with the problem or the moment that made the project click.
  2. (0:30–1:30) What you built. Show, don't tell. Live demo if possible.
  3. (1:30–3:00) How you used AI. Specific tools, specific prompts, specific moments where AI helped — and where it didn't.
  4. (3:00–4:00) What you learned. Not "I learned a lot." Specific, weird, hard-won insights.
  5. (4:00–5:00) What's next. The one thing you'd improve, the one thing you're now curious about.
What to Watch After Week 8

The field moves fast. These are the inputs worth keeping in your weekly diet:

The Mindset to Carry Forward
  • Use AI on something real every week. Atrophy is real. A skill you don't use weekly fades in a month.
  • Always verify, even when you "trust" the answer. Especially then.
  • Ship, don't lurk. The people who get good at AI fastest are the ones building visible things, not the ones reading takes.
  • Stay curious about the failures. A weird wrong answer is a window into how the model actually works.
"That's all, folks." — every cartoon. Except in your case, it's actually "that's all of this, folks." The AI story is just getting started, and you now have a place to stand inside it.
Going Deeper (Optional)

Tips for Getting the Most Out of This Curriculum

You don't need to be an AI expert. This curriculum is designed to learn by doing. Many of the exercises work great with a study partner, but they're equally effective working solo — compare approaches with a friend, classmate, or parent if you can.

Focus on the discussions. The exercises build skills, but the discussion questions build judgment. AI fluency isn't just knowing which buttons to press — it's knowing when to trust the output, when to push back, and when to do the thinking yourself. Those conversations are the most valuable part.

Embrace the struggle. Resist the urge to give up when a prompt doesn't work. The process of trying, failing, and iterating is exactly how prompt engineering skill develops. A bad result is a learning opportunity, not a dead end.

Make it real. If the capstone project involves a real business idea, a real community problem, or a real audience — even a tiny one — the motivation and learning multiply dramatically. An email newsletter with 12 subscribers is infinitely more educational than a hypothetical one with none.

Resource Hub

Free tools and references used throughout the curriculum

💬
ChatGPT
General AI assistant
🧠
Claude
AI for writing & analysis
Google Gemini
AI with massive context
🐋
DeepSeek
Chinese AI — compare & contrast
🔍
Perplexity AI
Research with citations
📓
NotebookLM
AI-powered note synthesis
📊
Gamma AI
AI presentation builder
🎨
Canva
Design & visual content
Zapier
No-code automation
📝
Notion
Workspace & project mgmt
🎙️
Otter.ai
AI transcription
✍️
Grammarly
Writing assistant
🌐
Carrd
Free one-page websites