L1–L6 All Galaxies
If AI could take over every part of your work, what would you spend your time on?
Every week, the gap grows exponentially.
Always exploring, always going in circles
What you saved isn't read yet, and new alerts keep coming.
Resources are everywhere, but you keep getting stuck.
A bit here, a bit there, it never connects.
Classes done, videos watched, still not sure you've got it.
After all this exploring, you need a complete learning system.
The answers you're looking for
What people lack isn't ability.
It's the fundamentals, and real use cases.
Here, you'll see how the front line puts AI to work.
A 15-minute AI interview finds the gap to close first and shows how well you use AI today.
Six lasting AI skills, built around real output, so you master the fundamentals step by step.
Practitioners share what really works, with real use cases and hands-on methods.
Swipe through the three cores →
From fundamentals to the front line, the answers are here.
All six skills, already planned and scheduled, so you don't build it yourself.
AI sounds convincing but misses your goal. You'll give it context and standards, and spot where the work drifted.
Your idea stays stuck in the chat. You'll turn it into articles, images or small tools with AI.
Every day you open AI, paste data, repeat. You'll hand off the whole job to run daily, and just check the results.
AI is fast, but too much output to review. You'll set checkpoints, read quality at a glance, and know the cost.
AI built a prototype, but it's not ready for others. You'll test it, protect the data, and ship it.
AI is hit or miss, and prompt tweaks are hard to judge. You'll set standards, find the failing step, fix it, and retest on the same cases.
Every skill you need, ready for you.
From single tasks to agents to whole systems, six AI fundamentals put you in charge. Take AI through the whole job.
L1–L2Get work done with AI
L3–L4AI works reliably for you
L5–L6Scale up and keep improving
AI Collaboration, AI Creation, Agent Design, AI Workflow Design, AI Product Development and AI System Optimization: one per galaxy.
In every galaxy you build real work with your teacher, from proposals and social graphics to a live product.
Built on the six-level AI skill standard, from L1 to L6. Every step sets up the next.
First get work done with AI, then let AI work reliably for you, then scale up and keep improving.
Skills with structure, practice with a path.
Wherever you're stuck, follow someone who's been there.
AI keeps getting better. What's left for me?
Break down your daily work, find what to give AI, and where your judgment is needed.
What if AI's code gets messier with every change?
Write the spec and split the tasks for AI to follow, then save versions with Git so you can recover when something breaks.
How do I turn one research piece into content for different platforms?
Build a content master from one piece of research, then write specs for each platform and a 60-second video script.
Going all-in on AI? What should leaders decide first?
From team alignment and process to business model, pricing and budget, settle the five decisions for bringing AI into a company.
Will AI-written code leak your secrets?
From protecting API keys and permissions to pre-launch checks, learn the security basics every AI product needs.
AI added a feature, and an old one broke?
Write tests for core features, review the code, and rerun everything after each change to catch what broke.
17 practitioners, each bringing what works on the front line today.
Brief clearly, make work you're proud to show
Your assistant starts daily, and handoffs don't break
Products people use, better with every version
Every role has a clear path to it.
Take homeChecked proposal, collaboration playbook, pre-upload data handling
Take homeArticles, video scripts, social graphics, small tools, interactive pieces, your own Skill
Take homeDaily intel assistant, four proofs it really ran
Take homeThree-color flowchart, three-stage checks, cost estimate, support workflow demo
Take homeExpense app, booking site, lookup tool, nine-point launch checklist
Take homeScoring rubric, failure log, fix log, retest results
After each part, you can see yourself change.
Hai Tseng
Waki
Carmen HsuehfOx Hsiao
Hans Lin
Dahao Wang
Vivi Chen
Karen
Eugene Wang
Eric Wu
Peggy LoRichard Lee
Wisely ChenEddie KaoAllen Own
Sean LiuSting Tao17 teachers · six galaxies
A new tool launches
Your manager hands you a task
Monday morning
Someone asks, "Can this go live?"
When AI gets it wrong
You know where it fits and whether it's worth switching. Decided in ten minutes.
You know what to give AI and what to judge yourself, and check every line before you hand it in.
Your assistant has this week's intel ready, and one research piece becomes this week's posts.
You can answer, and you can make it happen.
You know which layer failed, fix it, and retest on the same cases.
Swipe through the five moments →
Six galaxies, 35 classes.
With 6 AI assessments, see yourself change.
Sort out which work suits AI and make your own reference chart for every handoff.
Hai Tseng
Move from asking to delegating: pick your first task and refine it with assign, review, return until it works.
Waki
Use SCORE to state what you need and let AI ask questions first, so the first draft lands closer.
Waki
Write how you work with AI into a playbook, including how to prep files before upload, and take it to any task or tool.
Carmen Hsueh
Apply your playbook to a proposal, check figures and sources, and hand in verified work.
Carmen Hsueh
Turn your past writing into a writing spec and keep your voice on any new topic.
Hai Tseng
Build a content master from one piece of research, then write specs for each platform and a 60-second video script.
fOx Hsiao
Use the C.R.A.P. design principles to turn a key visual into a clear social post with editable text.
Hans Lin
Start from something you love and vibe-code a mini tool you'll actually use.
Dahao Wang
Add your own rules to a ready-made Skill and retest it with new data, so AI repeats your method.
Vivi Chen
Vibe-code four kinds of works with the instructor, from a front-end-only site to a tool that calls AI.
Karen
See what models, data, tools and agents each do, so you know what to give your assistant.
Eugene Wang
Use four criteria to pick a task for your assistant, map the flow, and grant only the access it needs.
Vivi Chen
Build a briefing assistant with ChatGPT and Notion that picks sources by your standards, remembers feedback and starts on its own every day.
Eugene Wang
Build a super assistant that prepares your morning briefing, writes up meeting notes and lines up your work each day.
Eugene Wang
Work around AI's two limits and draw a three-color rules/code/AI flowchart that gives each step the right method.
Eric Wu
Combine human review, AI self-checks and independent adversarial checks with a traffic-light rating to decide if the work can ship.
Peggy Lo
Write requirements as a spec AI can follow, let it break down the work, then accept it through tests and review.
Richard Lee
Work out the full cost of one task, match the models, and set budget caps and stop rules.
Wisely Chen
Break down a customer-support workflow with the instructor, pick SQL, RAG, an LLM or code for each step, then use problem emails to find what breaks.
Eric Wu
See the full path from spec through development and testing to launch, and spot what your project still needs.
Hai Tseng
Write the spec, split the tasks, let AI build to it, then check each item.
Eddie Kao
Save versions and branch with Git, so you can recover when something breaks.
Eddie Kao
Limit what AI can query, understand what each field means, and check the results.
Eric Wu
Protect keys and permissions and run a security check before launch.
Allen Own
Write tests for core features and rerun them after every change, so old features keep working as you add new ones.
Richard Lee
Deploy to a URL others can open, with login and test payments.
Eric Wu
Take a small product from spec to launch with the instructor, and have a second person use it.
Hai Tseng
Turn “good” into acceptance criteria, cap rounds and budget, and let AI revise until it passes.
Hai Tseng
Use good and bad examples with a rubric to write your taste as criteria AI can read.
Sean Liu
When AI gets it wrong, find whether facts, SOP, Skill or settings caused it, fix that layer, and retest with the same scorecard.
Carmen Hsueh
Separate read, write and irreversible permissions, cap costs, and hand high-risk actions to a person.
Wisely Chen
Trace a run to the step where it went wrong, and add that error to your test cases.
Sean Liu
Change one criterion, compare before and after on the same cases, then adopt or roll back.
Hai Tseng
Learn the five decisions to settle before bringing AI into a whole company.
Sting Tao
No. Each galaxy stands on its own. Start with the one you need most.
Take one before you start to see what to work on first, then again after each galaxy.
12 months, counted separately for each class.
Everything goes live on 2026/11/30 at 17:00.
Watch on your phone; build on a computer.All six galaxies can be watched on a phone, tablet or computer. Foundation lessons 01–03 can be practiced with an AI app on your phone; for the other galaxies we suggest a computer. In Twin, you set up on a computer and check results on your phone; Product requires a computer that lets you install software. Computer requirements: web-based AI just needs an up-to-date browser; Claude Code needs macOS 13 or Windows 10 (version 1809) or later and at least 4GB of RAM; the Codex desktop app needs an Apple silicon (M-series) Mac. See each galaxy's FAQ for details. To watch lessons, use Chrome 111, Safari 16.4 (iOS 16.4), Firefox 128 or later.
The method carries over; tools can change.Lessons mainly demonstrate Claude and ChatGPT, and the methods carry over when you switch tools. Foundation lesson 04 and Product use Claude Code or Codex; one is enough. Twin uses Claude or ChatGPT with Gmail, Google Drive or Notion; Workflow uses AI from different providers to check each other; Evolve uses Notion or another shared document tool.
Some free; Product needs paid tools.Foundation lessons 01–03 work on free plans; for lessons 04 and 05, Claude Code requires Claude Pro or higher, while Codex lets you start on ChatGPT's free trial allowance. Maker, Workflow and Evolve can start free, but rounds of revisions and re-testing use a lot of allowance, so we suggest a paid plan for at least one tool. Twin's scheduling and connectors may require a paid plan, which we recommend. Product requires paid tools, and hosting, domains and payment services for launching may cost extra. Check each tool's official website for current plans and limits. Whenever a lesson uses a paid feature, we'll flag it.
Want your whole team to join?
Learn about team plans →
