You don't need a $2,000 course. You need a few good free sources and about 100 hours of actually using the tools. Here's the shelf I'd point a beginner to, and what each one is good for.
Takes 10 minutes to set up. You need: nothing you don't already have. You end with: two sources picked, one project running, and permission to ignore everything else.
The shelf
1. The makers' own docs. Anthropic's prompting guide (for Claude) and OpenAI's help center (for ChatGPT) are free and written by the people who built the tools. Best for: learning how to actually talk to AI so you get usable answers. Start with: their guidance on giving context and examples. It's the highest-leverage hour on this whole page.
2. Google's free AI basics. Google publishes free "getting started with AI" material through Grow with Google. Best for: total beginners who want plain-language explanations, no jargon. Start with: one intro module, then stop and go use the thing.
3. Perplexity. Free to use, and it shows its sources. Best for: learning by asking real questions and seeing where the answers come from, a research habit, not just an answer machine. Start with: a question you'd normally Google, and click through to two sources to see whether it read them right.
**4. Two creators in your niche. Not ten. Pick two people who teach AI for your world (small business, your job, parenting) and follow only them for 90 days. Best for: cutting the noise so you go deep instead of chasing every new tool. Start with:** unfollowing eight. That's the part that actually works.
5. One newsletter, one podcast. A single weekly email and one podcast keep you current without living in the feed. Best for: staying updated in 20 minutes a week instead of 2 hours a day.
6. Your own "what worked" notebook. One doc where you paste every prompt that gave you a great result. Best for: building a personal library you reuse forever, this ends up more valuable than any course.
7. A real project you care about. Pick one task you do every week and rebuild it with AI. Best for: everything. You learn more from automating one annoying task than from 10 tutorials.
The power move: make the tool teach you
The best free AI teacher is the AI. Not for facts — for practice, because it can watch you work in a way a course never will.
Ten minutes of that beats an hour of watching someone else prompt. You're getting feedback on your own work instead of admiring theirs.
What a "what worked" entry looks like
Item 6 is the one people nod at and never do, usually because they think it needs to be organized. It doesn't:
2026-08-14 - client recap emails
Worked: gave it my notes AND last month's email so it
matched how I already talk. Told it "no bullet points."
Didn't work: asking for "professional" - got stiff.
Reuse: paste notes, add "match the voice of the example."Four lines, thirty seconds. Twenty of these and you have something no course could have sold you, because it's about your business and your voice.
What to skip, and why
Being honest about this is most of the value:
- Prompt packs with 500 prompts in them. You will use four. You'd have written those four yourself in ten minutes, and you'd understand why they work.
- Courses that teach a specific tool's buttons. The buttons change every few months. The thinking doesn't.
- Anything that leads with how much money you'll make. Watch what someone teaches you to do in the free content. If the free content is all promise and the how is behind a payment, that's the whole review.
- The urge to try every new tool. There will be a new one next week. The person who went deep on one is further ahead than the person who opened nine.
A quick filter for whether someone's worth learning from: do they show you the actual prompt? People who teach show their work. People who sell show results.
Your first two weeks
- Week 1: pick two sources off this shelf, and unfollow the rest. Start the notebook, even if it has one entry.
- Week 2: pick the task you dread and rebuild it with AI, using the prompting habits from source 1. Add what worked to the notebook.
That's it. Two sources, one project, one notebook. If you did nothing else for two months you'd be ahead of most people who bought the course.
When it goes sideways
"I'm collecting, not doing." The tell is a folder of saved posts and no changed tasks. Close everything and do item 7 today.
"Everything sounds urgent and I can't keep up." You can't, and neither can anyone selling you that feeling. The tools get easier, not harder. Something you skip today will be simpler when you get to it.
"I can't tell good advice from hype." Use the filter above, and add one test: ask the AI itself. Paste in the advice and ask "what would have to be true for this to work for a business like mine, and what does it leave out?"
"I don't have 100 hours." Nobody does, in a block. It's twenty minutes a day for a year, most of it inside work you were doing anyway. That's the whole point of item 7.
Do this next
Pick two. Unfollow the rest. Open a blank doc, title it "what worked," and put today's date in it.
The only rule: depth beats collecting. Start tonight with the task you dread most — the One-Task Method guide is the shortest path from here to something actually working.