Each academy is two projects. The first one teaches the move. The second one makes it real. By the end of the second project, you have something a real person can use.
The first academy. Two projects: in the first you write a Claude Skill that knows what only you know. In the second you make it actually work — you watch friends try it, fix what confused them, and teach it when to say "I don't help with that." By the end you have a Skill someone in your life uses.
Pick a thing you know more about than your friends. Write a Skill that thinks the way you do about it. By the end, the Skill produces something a generic prompt to Claude could not have.
Show your Skill to 3 friends. Watch where they get stuck. Make 3 fixes. Add a "what I don't help with" section so it knows when to stop trying.
Real HTML, CSS, JavaScript. No drag-and-drop. No platform lock-in. By the end of project two, you have a real web page on a real URL that does something neither you nor Claude could have built alone. No API keys, no servers — Claude joins your page through a clean copy-paste workflow that's safe and simple.
HTML / CSS / JavaScript in plain text files. A real URL by the end (Netlify Drop, free). Built for one person you can name.
Use Claude in claude.ai to generate the smart parts of your page. Save the output as a JSON file. Your page reads it. No API keys — but Claude is doing real work.
The most empathetic academy. An agent isn't a chatbot — it's something that does things for a person. Project one: pick a real person, watch them, and design an agent for what they actually need. Project two: figure out the rules — what should the agent never do, even when asked? Both projects end in a written design, not a built agent (that's for high school).
Pick someone real. Watch them for 3 days. Take notes on what they actually struggle with. Write the agent spec — what it would do, what it wouldn't. No code yet.
Take your spec. Add 3 permission tiers (suggest / ask / do). Write 3 disagreement scenarios — what happens when the user asks for something the agent shouldn't do?
The smallest academy, and the most personal. Project one: pick something you make weekly — a journal entry, a sketch, a piece of writing — and define what makes a good one. Three dimensions, three anchors. Project two: score yourself for three weeks and see if you're drifting toward AI's average or toward your own taste.
Pick a domain you care about. Write a 3-dimension rubric with anchors. Score 3 examples by hand. Notice what your taste actually is when you have to write it down.
Run your rubric on your own work, weekly, for 3 weeks. Read the trend line. Make one fix. Write a half-page reflection on what changed and why.