What it is. Who makes it. How to use it — and how to put it to work for your family, your work, and yourself.
What you'll walk away with
─────────────────────────
✓ Mental model for AI
✓ Know your tools
✓ Understand agents & bots
✓ A real app you built
Format: ~2 hours
Level: No code required
Instructor: Regina Flores Mir
── reginaflores.ai ──
scroll to explore
The Map
AI is not new
AI is the big umbrella. Machine Learning is how we build AI today. Inside ML are three learning approaches — and inside one of them (supervised learning) lives the neural network stack that eventually gave us Claude.
Supervised Learning
Learns from labeled examples. You show it thousands of emails tagged "spam" or "not spam" and it learns the pattern. Most practical AI is this.
No labels — the model finds hidden patterns on its own. You don't tell it what to look for. It discovers structure in the data you didn't know was there.
🎵 Spotify grouping songs by mood
🛍 Amazon "customers like you"
🧬 Finding cancer subtypes in patient data
💳 Flagging unusual spending patterns
Reinforcement Learning
Learns by trial and error, earning rewards for good decisions and penalties for bad ones. No labeled data — just feedback from the environment.
♟ AlphaGo beating world chess champion
🤖 Teaching robots to walk
🎮 AI mastering video games
✦ RLHF — makes Claude safe + helpful
Classical software is explicit: a programmer writes rules. ML inverts this. You define an objective, feed in data, and an algorithm adjusts millions of internal parameters until the model learns to minimize its error. No rules are written. Behavior emerges from optimization.
01 — Data
Feed it examples
Billions of text passages, images, code snippets, and conversations become the raw material for learning.
02 — Objective
Define "good"
A loss function measures how wrong the model is. The goal: get that number as low as possible, as reliably as possible.
03 — Optimize
Adjust billions of weights
Gradient descent iteratively tunes parameters. Run this long enough at sufficient scale and complex capabilities emerge.
04 — Align
Shape the personality
RLHF — Reinforcement Learning from Human Feedback — turns a raw predictor into something helpful, safe, and pleasant to talk to.
The Two Breakthroughs
ML existed for decades. These two things changed everything.
Modern AI didn't come from one invention. It came from two distinct advances — one in architecture, one in alignment — that together turned raw prediction into something that feels like intelligence.
2017 — Google
Transformer Architecture
Before the Transformer, AI read text sequentially — word by word, left to right. The breakthrough was attention: letting the model look at all words simultaneously and learn which ones matter most to each other.
In the sentence "The animal didn't cross the street because it was too tired" — attention is what tells the model that "it" refers to "animal," not "street." That ability to track relationships across context is what makes LLMs feel like they understand.
Vaswani et al., "Attention Is All You Need," 2017
2022 — OpenAI
RLHF
A trained model is technically just a very good text predictor. It doesn't yet know how to be helpful or safe. Reinforcement Learning from Human Feedback is the fine-tuning step that shapes it into an assistant.
Human raters compare pairs of model responses and pick the better one. Those preferences become a reward signal, and the model learns to produce outputs humans rate highly. This is what turns a raw predictor into something that follows instructions, avoids harm, and feels like it has a personality.
Why Claude feels different from GPT — same architecture, different RLHF.
One Method, Many Architectures
ML is the umbrella. The models beneath it vary wildly.
The same optimization principle underlies completely different model families — each with a different architecture, training signal, and class of problems it solves well.
Most AI tools you use today combine more than one. ChatGPT and Claude are LLMs fine-tuned with RL. DALL·E is a diffusion model guided by a text encoder. Sora combines both. (Sora shutting down Apr 26)
💬
Large Language Models
Transformer-based. Trained to predict the next token across text, code, and data.
ClaudeGPT-5Gemini
🎨
Diffusion Models
Trained to reverse a noise process. Start with static, iteratively denoise into an image or video.
MidjourneyStable DiffusionSora (shutting down Apr 26)
🏆
Reinforcement Learning
Trained via reward signals from an environment. The model discovers strategies by trial and error.
AlphaGoRoboticsRLHF fine-tuning
👁
Vision Models
Trained on images for classification, detection, and segmentation. Vision Transformers now lead on benchmarks.
Face IDMedical imagingSelf-driving
🌐
Multimodal Models
Combine architectures to handle text, images, audio, and video together. The current frontier.
GPT-5Gemini 3.1Claude Opus 4.6
Why this matters
When someone says "AI did this" — a song, an image, a diagnosis, a recommendation — a different model family is behind each one. They share the same mathematical foundation but are built for completely different problems.
The Landscape
AI isn't a chatbot — it's everything
The tools you consciously choose are just the tip of the iceberg. AI has been quietly running your daily life for years — you just didn't call it that.
💬Text & ChatClaude · ChatGPTGemini · Perplexity
🎨Image GenMidjourney · DALL·EFirefly · Ideogram
🎬VideoSora (shutting down Apr 26) · RunwayPika · Kling
Given a prompt, generates something that didn't exist before — text, images, music, video, code.
💬 Claude · ChatGPT🎨 Midjourney🎵 Suno🎬 Sora (shutting down Apr 26)
03 — Agentic
Takes action on your behalf
Goes beyond answering — it actually does things. Opens apps, browses the web, writes files, sends emails, completes tasks end-to-end.
💻 Claude Code📧 AI email handler🗓 Calendar booker🛒 Auto-orders
The blurring line
Most tools today combine all three. Claude is generative when you chat with it, and agentic when you use Claude Code or give it tools. The lines are disappearing fast — and the agentic layer is where everything is heading.
The Players
Who's building the major AI models?
There isn't one AI — there are several, made by competing companies with different personalities, strengths, and philosophies. Here's how to think about the main ones.
🤖
OpenAI
ChatGPT · GPT-5
The company that started the public AI wave with ChatGPT in 2022. Backed by Microsoft. The most recognized name — most people's first AI.
KNOWN FOR
Widest brand recognitionBest all-rounderCanvas editor for writingDALL·E image gen
LATEST MODEL
GPT-5.3
$20/mo · ChatGPT Plus
we use this
✦
Anthropic
Claude · Sonnet · Opus
Founded by former OpenAI researchers focused on AI safety. Claude is known for being thoughtful, nuanced, and exceptionally good at writing, reasoning, and long documents.
KNOWN FOR
Best writing qualitySafest outputsClaude Code + Cowork1M token context
LATEST MODEL
Claude Opus 4.6 · Sonnet 4.6
$20/mo Pro · $100/mo Max
🔷
Google DeepMind
Gemini · Gemini Advanced
The search giant with the deepest AI research history. Gemini is built into Google Search, Gmail, Docs, and more. If you use Google products, you already have access.
KNOWN FOR
Best benchmark scores right nowGoogle Workspace built-inBest value pricingVideo + audio
LATEST MODEL
Gemini 3.1 Pro
$19.99/mo · Google One AI Premium
🔵
Meta AI
Llama 4 · Open Source
Facebook's parent company took a different bet — releasing their models as open source. Anyone can download and run Llama. It powers thousands of other products and apps.
KNOWN FOR
Free and open sourceRuns privately on your devicePowers hundreds of products10M token context
LATEST MODEL
Llama 4 Maverick · Scout
Free to download · No subscription
𝕏
xAI (Elon Musk)
Grok 4 · Real-time X data
Built into X (Twitter). Grok's edge is access to real-time posts and news — it knows what happened in the last hour. Also the cheapest frontier model by API cost.
KNOWN FOR
Real-time X dataCheapest API costsLeads coding benchmarksFewer content restrictions
LATEST MODEL
Grok 4
$22/mo · X Premium+
🌊
DeepSeek
China · Open Source · R1
The Chinese lab that shocked the industry in early 2025 by releasing a model that matched GPT-4 at a fraction of the cost. Raised serious questions about the US lead in AI.
Gemini is Google's paid product. Gemma is Google's free, open-source cousin — a lighter model you can download and run privately on your own device. Same research lineage, no subscription required.
KNOWN FOR
Free to downloadRuns on your own hardwarePrivacy — no data sent to GoogleGoogle research quality
THINK OF IT AS
Google's version of Meta Llama
Free · Developer / researcher tool
🔵
Alibaba — Qwen
China · Open Source · Multilingual
Alibaba's open-source model family — and one of the biggest surprises in AI. Qwen 3.5 is now genuinely competitive with GPT-4 class models and is dominant in multilingual and non-English tasks.
KNOWN FOR
Best multilingual modelOpen source90,000+ enterprise adoptersClosing gap fast
LATEST MODEL
Qwen 3.5
Free · Alibaba Cloud API
Mistral AI
France · Open Source · Fast
The European challenger. French AI lab building open-source models known for speed and efficiency. Mistral punches well above its weight — smaller models that outperform much larger ones.
KNOWN FOR
Best speed-to-quality ratioEuropean / GDPR-friendlyGreat for codingTruly open source
LATEST MODEL
Mixtral · Magistral 2
Free · mistral.ai
CLOSED vs. OPEN
Closed models (OpenAI, Anthropic, Google) keep their code private. Open models (Meta, DeepSeek) publish it — anyone can build on top, modify, or run them privately.
THE BOTTOM LINE
There's no single winner. Each model has a personality and a sweet spot. Most power users keep 2-3 open and switch depending on the task.
WHY WE USE CLAUDE
Best writing quality, strongest safety track record, and the most complete agentic ecosystem — Code, Cowork, Chrome. The most complete platform for building real things.
NOTABLE EXCEPTION — MICROSOFT
Microsoft doesn't compete with a frontier model of its own. Instead they invested $13B in OpenAI, struck a $30B compute deal with Anthropic, and built Copilot on top of both. Their open-source model Phi 4 is small but excellent. Their strategy: be the platform and infrastructure for all the best models — not build one themselves.
Going Deeper
What is a Local LLM?
Normally when you use Claude or ChatGPT, your words travel to a server, get processed, and come back. A local LLM runs entirely on your own computer. Nothing leaves your machine.
CLOUD AI (what you normally use)
→
Your message goes to Anthropic's / OpenAI's servers
→
Processed in the cloud, response sent back
→
Most powerful models available
→
Requires internet · Monthly subscription
→
Company can see your prompts (per their privacy policy)
LOCAL LLM (runs on your computer)
✓
Everything stays on your machine — zero data sent out
✓
Works offline — plane, no wifi, anywhere
✓
Download once, use forever — no monthly fee
✓
No usage limits, no rate throttling
↓
Less powerful than cloud frontier models — you're trading raw capability for privacy
WHEN DOES THIS ACTUALLY MATTER — FOR YOUR LIFE
⚕️
Medical records
Analyzing health documents, insurance paperwork, or a family member's diagnosis — locally, privately.
⚖️
Legal documents
Contracts, estate planning, NDAs — you want AI help but you don't want those words on anyone's server.
💼
Confidential work
Board memos, M&A documents, client strategy — anything with NDA implications or fiduciary responsibility.
👧
Your kids' information
School records, IEP documents, pediatric notes — some parents prefer this stays entirely off cloud servers.
HOW TO RUN ONE — EASIER THAN YOU THINK
01
Download Ollama
Free app for Mac and Windows. Installs in 60 seconds. ollama.com
02
Pull a model
Type one command: ollama run llama4 and it downloads automatically.
03
Add a friendly UI
Apps like LM Studio give you a ChatGPT-style interface — no command line needed at all.
WHAT RUNS WELL LOCALLY
Meta Llama 4 ScoutBest overall
Google Gemma 3Lightweight · Fast
Microsoft Phi 4Best on small devices
MistralFast · Multilingual
THE HONEST TRADEOFF
A local LLM is like having a very capable assistant who works entirely inside your house and never tells anyone anything. They're not quite as brilliant as the best cloud models — but they're completely yours.
Specialization
What is a Vertical LLM?
A local LLM is about where it runs. A vertical LLM is about what it knows. These are models trained specifically on one industry's language, documents, and data — making them far more precise than a general model in their domain.
GENERAL LLM
Knows a little about everything. Great for most tasks. Like a very well-read generalist.
VERTICAL LLM
Trained deeply on one field. Speaks the language. Knows the nuance. Like a specialist with 20 years in the room.
REAL EXAMPLES BY INDUSTRY
⚖️
Legal
Harvey
Trained on case law, contracts, and legal precedent. Used by major law firms to review documents and draft arguments.
📈
Finance
Bloomberg GPT
Trained on 40 years of financial documents, earnings calls, and market data. Understands financial language at a level general models can't match.
⚕️
Healthcare
Tempus · Med-PaLM
Trained on clinical records, oncology research, and medical imaging. Helps doctors analyze diagnoses and treatment options.
🎨
Creative
Adobe Firefly
Trained exclusively on licensed creative assets — no copyright issues. Built specifically for designers and creative professionals.
🏫
Education
Khan · Khanmigo
Built on Khan Academy's entire curriculum. Acts as a personal tutor — never just gives the answer, guides the student to find it.
🛍️
Retail & Fashion
Jasper · Stylist AI
Trained on brand voice, product catalogs, and consumer behavior. Writes marketing copy and product descriptions that convert.
THE MATRIX
CLOUD
LOCAL
GENERAL
Claude, ChatGPT
Llama, Gemma
VERTICAL
Harvey, Bloomberg GPT
Rare — enterprise only
WHY IT MATTERS TO YOU
If you ever work with a specialized AI tool in your field — a legal research assistant, a financial planning tool, a tutoring app for your kids — you're likely using a vertical LLM under the hood. The general models you're learning today are the foundation everything else is built on.
The Anthropic Product Stack
One company. A whole ecosystem of tools.
You've heard of Claude. But Anthropic has built an entire suite of products — each one designed for a different level of access and a different kind of user. Here's how they all fit together.
LAUNCH TIMELINE
Nov 2024
Claude Code (CLI)
Terminal-based coding agent. Developers only — but people immediately started using it for everything else.
Oct 2025
Claude Code Web Interface
No terminal required. Claude Code moves into the browser and then into a downloadable Desktop App.
Jan 13, 2026
Claude Cowork launches ✦
Claude Code for non-developers. Mac only, Max plan only at launch. Lets Claude read and edit files in a folder you designate — no terminal needed.
Feb 2026
Cowork expands: Plugins + Pro access
Google Workspace, DocuSign, and other enterprise connectors added. Pro plan subscribers get access. Scheduled recurring tasks launch.
Apr 3, 2026
Cowork + Computer Use → Windows ← we are here
Full Windows parity. Computer use: Claude can now see your screen, click, type, and navigate apps on your behalf. Pro + Max.
THE FULL PRODUCT LINEUP
💬
claude.ai
Claude Chat
The conversational interface. Web + mobile. Free tier available. Where most people start.
FREE · PRO · MAX
💻
Desktop App
Claude Code
Builds apps, writes code, manages files. Works in a browser window or the Desktop App. No terminal required.
PRO · MAX · ENTERPRISE
🗂️
Desktop App · NEW
Claude Cowork
Claude Code without the code. Designate a folder → Claude reads, edits, and creates files on your behalf. Can now control your screen.
PRO · MAX · Research Preview
🌐
Chrome Extension
Claude in Chrome
Browses the web on your behalf. Research, summarize, fill out forms — all from the browser sidebar.
PRO · MAX
📊
Add-ins
Claude for Excel · PowerPoint
Claude lives inside your Office apps. Analyze data, build slides, and format documents without copy-pasting.
PRO · MAX · ENTERPRISE
🔌
Connectors / Plugins
Google, Slack, DocuSign…
Claude connects to the apps you already use. Works inside Cowork and Claude Code. Expanding fast.
VARIES BY PLAN
COWORK IN PLAIN ENGLISH
Think of it as hiring a very fast, very thorough assistant who works inside your computer.
You point Claude at a folder. You give it a task. It reads your files, figures out what needs to happen, does it, and tells you when it's done. You can walk away. No babysitting required.
EXAMPLE TASK
"Read all the PDFs in my Downloads folder and make me a spreadsheet with vendor, amount, and date for every receipt."
EXAMPLE TASK
"Every Monday morning, scan my email, summarize anything that needs my attention, and draft replies for the ones that are routine."
EXAMPLE TASK
"Take all the photos from our ski trip, rename them by date and location, and put together a simple photo book layout."
The bottom line
Claude Code is for building things. Cowork is for getting things done. If you've ever wished you had an extra set of hands for the administrative layer of your life — scheduling, organizing, researching, formatting — Cowork is what that looks like.
Under the Hood
The three Claude models
Whether you use Claude in the browser, the desktop app, or the terminal, you're talking to one of three underlying models. Think of them as different settings on the same engine.
Most Powerful
Opus 4.6
The deepest thinker. Best for complex reasoning, nuanced writing, and hard problems where quality matters more than speed.
Complex analysisResearchLong documents
Best Balance · What you'll use most
Sonnet 4.6
The everyday workhorse. Fast, smart, and capable — handles the vast majority of tasks beautifully, including today's workshop.
CodingWritingDaily tasksAgents
Fastest
Haiku 4.5
Lightning-quick for simple tasks. Great when you need a fast answer, a short summary, or a quick rewrite on your phone.
Quick answersSummariesSpeed
These three models power every way you access Claude — the website, the mobile app, the desktop app, and the terminal. The interface changes; the intelligence underneath doesn't.
Announced today — April 7, 2026
Meet Claude Mythos
Anthropic just unveiled its most powerful model yet — one so capable they're not releasing it to the public. This is what the frontier of AI looks like right now.
WHAT IT IS
Above Opus — a new tier entirely
Anthropic has always had Haiku → Sonnet → Opus. Mythos is a new category above all of them — dramatically more powerful than anything they've released before.
Found thousands of security holes — in hours
Mythos scanned major software systems and discovered tens of thousands of critical bugs — some of them decades old and never caught by human researchers. It then wrote the fixes.
Too powerful to release publicly — for now
Anthropic is so concerned about what Mythos could do in the wrong hands that they're keeping it invite-only. Only ~40 companies in the world have access.
PROJECT GLASSWING — WHO HAS ACCESS
Apple
Amazon
Microsoft
Google
JPMorganChase
CrowdStrike
Cisco
Nvidia
+ ~32 more organizations. No public access. No timeline for general release.
$100M COMMITTED
Anthropic is providing up to $100 million in usage credits to companies testing Mythos — plus $4 million directly to open-source security foundations.
WHY THIS MATTERS TO YOU
The apps on your phone. Your banking software. The browser you're using right now. Mythos is scanning all of it for holes that hackers could exploit — and patching them before anyone gets hurt. This is AI working for you in ways you'll never directly see.
THE BIGGER SIGNAL
We are now in a world where AI models are so capable that the companies building them are pausing before releasing them. That is a new thing. The question of what AI should be allowed to do is no longer hypothetical.
A test so hard that when AI finally aces it, we may no longer be able to write one harder. 2,500 questions from nearly 1,000 subject-matter experts at 500+ institutions. In early 2025, the best models scored under 10%. By March 2026, GPT-5.4 Pro leads at 44.3% — human experts still average 90%.
WHAT IT IS
The problem: AI was acing every test we threw at it. MMLU, the gold standard benchmark, is now solved at 90%+ accuracy. We ran out of hard tests.
The solution: Ask the world's sharpest minds to submit their hardest questions. PhD-level. Verifiable answers. No multiple choice tricks. Published in Nature, January 2026.
The result: Even the best models struggle. GPT-5.4 Pro leads at 44.3%. These are questions that genuinely stump frontier AI.
MODEL ACCURACY ON HLE — Scale AI Official Leaderboard · March 2026
GPT-5.4 Pro
44.3%
Gemini 3 Pro Preview
37.5%
GPT-5.4 (thinking)
36.2%
Claude Opus 4.6 (thinking)
34.4%
GPT-5 Pro
31.6%
GPT-5.2
27.8%
GPT-5
25.3%
Claude Opus 4.5 (thinking)
25.2%
Claude Opus 4.6 (no thinking)
19.0%
0%
25%
50%
75%
100%
Source: labs.scale.com/leaderboard/humanitys_last_exam · March 2026 · Human expert avg: ~90%
WHY IT MATTERS
When a model hits 50%+ on HLE, it means AI has reached expert-level reasoning across essentially every field of human knowledge. We're not there yet — but the pace of improvement suggests we will be very soon.
BUT WAIT — HLE DOESN'T TEST EVERYTHING
70% of developers prefer Claude for coding.
Claude leads on what matters in real engineering — multi-file reasoning, accurate refactoring, fewer hallucinated APIs. Claude Code owns 54% of the enterprise coding market.
A New Way to Build
Vibe Coding
Coined by Andrej Karpathy — one of the founders of OpenAI — in early 2025. Instead of writing code yourself, you describe what you want in plain English, and the AI builds it. The skill isn't coding anymore — it's knowing what to ask for and recognizing when it's right.
The old way
Learn Python. Study HTML. Understand databases. Take courses for months. Write hundreds of lines of code. Debug errors. Start over. Repeat indefinitely.
Vibe coding
Open Claude Code. Describe what you want. Watch it appear in seconds. Give feedback. Iterate. Walk away with something real.
"The best interface for AI is the one you've been using your whole life — plain English."
The old skill
Prompt engineering
For a moment, people sold courses on how to word your inputs. Frameworks, formulas, incantations.
The new skill
Clear intention
Be specific. Give context. Say what you actually want. Push back when it's wrong. That's it. You already know how to do this.
Claude Code
You don't need a terminal. You don't need VS Code.
Claude Code used to be a terminal-only developer tool. Anthropic now offers a Desktop App that gives you the same full power through a regular window — download and open it like any other app.
Step 1
Go to claude.ai
Create a free account with your email address. No credit card required to start.
Step 2
Upgrade to Claude Pro
$20/month. This unlocks Claude Code. The free plan doesn't include it.
Step 3
Download the Desktop App
Available for Mac and Windows. Installs like any normal app — no terminal involved.
Step 4
Open and start talking
Describe what you want to build in plain English. No coding experience required. Just a conversation.
Web Apps vs. Native iOS Apps
Web apps run in any browser — phone, laptop, tablet — and can be added to your home screen as a Progressive Web App (no App Store, no fees). Native iOS apps require Xcode, a Mac, and a $99/year Apple Developer Account. For this workshop, we're building web apps — faster, free, and just as useful.
New
AI Agents & Bots
A chatbot answers your question. An agent goes and handles it. This is the most important shift happening in AI right now — and it's already available to you.
HOW AN AGENT WORKS
🧑💻You give itA Goal
→
🧠It plansA Strategy
→
🔧It usesTools & APIs
→
✅It deliversThe Result
CHATBOT VS. AGENT — THE KEY DIFFERENCE
Dimension
Chatbot
Agent
What it does
Answers your question
Completes your task
Memory
This conversation only
Can remember across sessions
Tools
Text in, text out
Browses web, reads files, sends email
Runs for
Seconds
Minutes to hours, unattended
You do
All the follow-up yourself
Review the finished output
REAL EXAMPLES — FOR YOUR LIFE RIGHT NOW
📅
School Scheduler
Tell it: "Parse the Buckley weekly email, pull all Grade 2 events, and send me a weekly digest every Sunday night."
Try with Claude Code
📧
Email Triage Bot
Reads your inbox, flags what needs a reply today, drafts responses for your review. You approve, it sends.
Try with Claude + Gmail
🔍
Research Agent
Tell it: "Research summer programs for a 7-year-old in NYC, under $3k, with spots still open." It searches, compares, and returns a ranked list.
Try with Perplexity
🛒
Shopping Agent
"Find the best-reviewed lunchbox under $40 that fits in a Buckley cubby and has no BPA." Done. With links.
Try with ChatGPT Operator
THE MINDSET SHIFT
Stop asking AI for answers. Start assigning AI tasks — with a deadline and a deliverable.
Workshop
Let's build something real.
In the next 30 minutes, we're going to use Claude — Chat, Cowork, and Code — to build a real, working app together. No coding experience needed. You just need to know how to talk.
30minutes to build
0lines of code you'll write
1real app you'll walk away with
The App
The Buckley School Smart Calendar
The school calendar PDF has everything — but it's a wall of text. This app turns it into a filterable, color-coded calendar so you can see exactly what matters to your child's grade. No more scrolling through dates that don't apply.
🎓
Filter by Grade
See only events relevant to your child's class. One toggle, one view.
🔴
Color-coded by type
Red = days off · Blue = school events · Green = deadlines
📅
Month view or list view
Toggle between a calendar grid and a clean scrollable list.
📲
Works on your phone
Add to home screen as a PWA. No App Store. No download. Free.
"Here is the Buckley School calendar PDF. Parse all the events, dates, grades, and types into structured data."
02
Build the calendar
"Create a web app that displays these events as a calendar. Clean design, easy to read on a phone."
03
Add filters
"Add a filter by grade so I only see events for my child's class. Add a second filter by event type — days off, school events, deadlines."
04
Make it beautiful
"Color code: red for days off, blue for school events, green for deadlines. Make it polished enough to share with other Buckley parents."
05
Add an agent layer
"Add a weekly digest feature: every Sunday, summarize next week's events by grade and format it as a message I can send to our family group chat."
This is a real app you'll actually use. When you see something built from your own school's data, it stops feeling like a demo and starts feeling like a superpower.
Workshop Steps — Continued
What you'll say to Claude Cowork & Code
Claude Cowork
01
Create a folder
"Create a new folder called Buckley Calendar on my desktop and set it as your Cowork workspace."
02
Add the file from Chat
"Save the structured calendar data we built in Claude Chat into this folder as calendar-data.json."
Claude Code
01
Push to GitHub
"Create a GitHub repository called buckley-calendar, commit all the files, and push. Then enable GitHub Pages so the app is live at a public URL I can share."
✦
By the end: a live web app at a real URL, a recurring weekly digest, and your first GitHub repo — all from plain English conversations.
You already have everything it takes.
Curiosity, taste, and knowing what you need — that's the real skill.
AI just gives you a way to act on it, faster and bigger than ever before. The version of you that knows how to use these tools is dramatically more capable than the version that doesn't.
