Somebody in your feed is selling a $497 "AI mastery" course this week.
Here's what stings. A big chunk of what's in it came from free material published by the companies that actually build these models. They didn't invent it. They watched the free thing, put a logo on it, and charged you.
I'm not mad that people sell courses. I sell things. I'm mad that people sell the free thing and count on you not knowing.
So here's the free version. What it is, who it's for, and the order that actually works.
Start here, not with a course
Before you enroll in anything... use it for a week on something real.
Courses teach you what the buttons do. They don't teach you what it feels like when the answer comes back confidently wrong and you have to catch it. That only comes from reps on your own work.
One real task from your own week beats 10 hours of tutorials. Every time.
Then take a course, because now you'll know which parts matter.
The free stuff, ranked by who you are
If you've never used it seriously
Elements of AI (University of Helsinki, free). No code. It explains what these systems actually are and what they are not, which kills about half the fear people carry around. A few hours, self paced.
Google's AI Essentials. Lives on Coursera. Aimed squarely at normal working people, not engineers. Coursera lets you audit the material for free and charges for the certificate, so decide whether you want the paper or the knowledge.
Anthropic's own learning material. The people who build Claude publish free courses on how to use it well. Straight from the source, no guru in the middle. That's the one I'd do first if Claude is your tool.
If you're past beginner and want to build
DeepLearning.AI short courses. Andrew Ng's outfit. Short, free, focused, taught with the actual companies who make the tools. Prompt engineering, agents, RAG, evaluation. This is the best free technical material on the internet and it isn't close.
Microsoft's "Generative AI for Beginners" on GitHub. Free, open, 18 or so lessons with real code you can run. It's a repo, not a slick platform, and that puts people off... it shouldn't. The content is strong.
Hugging Face courses. Free, deep, and the place to go once you want to understand what's under the hood instead of just driving it.
If you want the paper for your resume
Certifications are worth less than people think and more than nothing. They get you past a filter. They don't get you hired.
Google, Microsoft and a few of the cloud providers all run AI certificates, usually free to learn and paid to certify. If a job posting names one, get that one. If none do, skip it and build something you can show instead. A thing you made beats a badge, every time.
More on that in the resume guide.
Rest of it's yours
Drop your email and the whole thing opens up right here. No waiting on a confirmation, no download, nothing to install.
You'll also get the new ones as I make them. Leave whenever, no hard feelings.
The order that actually works
Most people quit in week 3. Not because it's hard. Because they never connected it to anything they actually do, so it stayed a hobby, and hobbies lose to a full week.
Here's the sequence that sticks:
- Week 1. Use it, badly. Pick one thing you do every week that you hate. Make it do that. It'll be rough. Fine.
- Week 2. Set it up properly. Tell it about you and your work once, so it stops giving you answers written for nobody. That's the 15 minute setup and it changes everything downstream.
- Week 3. Take one real course. Now the material lands, because you have somewhere to put it.
- Week 4. Build one thing you'd show somebody. Not a toy. A thing that saves you or somebody else actual time.
If you want that expanded day by day, the 30 day path is here.
YouTube: the good, and the trap
YouTube is genuinely the best free university that has ever existed for this. It's also 90% people re-reading a press release to you.
What to look for:
- They show the screen. The whole screen, including the part that breaks.
- They show the prompt, not just the result.
- Old videos still work. If everything on the channel is 6 days old and about the newest release, they're chasing news, not teaching.
- They tell you what it can't do.
What to run from:
- Thumbnails with a shocked face and a number in red
- "This changes EVERYTHING"
- Anybody whose entire channel is announcements
- Anybody who won't show a failure
That last one matters most. Watching somebody get it wrong and fix it teaches you more than 20 clean demos, because the fixing is the actual skill. The clean demos are edited.
Free courses from real universities
Harvard, MIT, Stanford and a pile of others put real coursework online for free. edX and the university sites host most of it. You audit for free, you pay if you want the certificate.
Harvard's CS50 line is the famous one and it deserves it. There's a general intro and an AI-focused follow on. It is a real course with real workload. Don't start there if you've never touched this. Do go there when you want depth.
The catch with university material: it's usually teaching you how AI works, not how to use it Tuesday morning. Both are worth having. Know which one you signed up for.
What none of the free stuff will do for you
It won't tell you what to build. That part's yours.
Courses hand you capability. They don't hand you a reason. The people who get somewhere with this all did the same thing: they picked one annoying, real, specific problem in their own life or work and refused to stop until it was handled.
The course is the easy part. Picking the problem is the whole job.
The short version
- Use it on real work for a week before you enroll in anything
- Do the free course from whoever built the tool you're using
- DeepLearning.AI for depth, free
- Elements of AI or Google's AI Essentials if you're starting cold
- Certificates only if a job posting names one
- On YouTube, trust the people who show the part that breaks
- Nobody's course will pick your problem for you
You don't need to spend a dollar to get good at this. You need reps and about 6 weeks.