Everyone loves to tell you what really matters when learning AI.
It's the tutorials.
No wait, it's the YouTube videos.
Actually, it's that one course that promises to take you "from zero to AI master in 7 days," taught by a guy standing in front of a blue-purple gradient.
It's none of those.
The most important thing is spending money.
I learned this the hard way, and now I can't unsee it.
The "ouch" number
Here's the rule: you need to spend an amount of money that genuinely makes you wince.
Not break-the-bank money.
Wince money.
Say you make $2,000 a month. Spend $1 on something and who cares? That's not a decision. It doesn't even register as a scratch.
But go the other way and YOLO the whole $2,000, and that doesn't work either. You've gone bankrupt before the course even loads.
The ouch number lives somewhere in the middle.
Around $200.
It won't break a bone, but it's enough that your chest does a little thump the moment you tap "pay." Enough to make it a real decision.
Which is, funnily enough, almost exactly what I'm currently paying for Claude Code Max. $200 a month.
I did not pick that number for this post. I'm literally writing this on the very subscription I'm describing, which is either great editorial discipline or a cry for help.
git blame me later.
That little thump is the whole point.
Your sunk cost needs to stand up
Normally, sunk cost is the villain of every decision-making blog post.
Here, it's the hero.
Because humans are simple creatures. As long as not learning costs nothing, we'll happily sit on the phrase "I want to learn this" forever.
"I want to learn AI."
"I want to learn to code."
"I want to learn video editing."
You can say that for a year and still write zero lines of code.
Why?
Because there is no sunk cost. Nothing pulling you back to the desk. Nothing quietly asking why you are watching another productivity video instead of actually becoming productive.
You have to make your sunk cost stand up.
The second the money leaves your account, it stands up and starts running in the background, poking you in the back:
You already paid $200, and you're still not doing it?
It's basically a cron job for your guilt.
And this isn't some hot take I'm inventing out of thin air. Almost everything you want to learn is already on YouTube for free. Full courses are sitting on Coursera. Documentation is everywhere. Tutorials are everywhere.
So why do people still buy the paid version — and often finish it more consistently?
Not magic.
The free version may be complete. The content may all be there. But people watch 10%, feel vaguely productive, and quietly close the tab because nothing happens if they do.
The paid course isn't always better.
You are more committed.
The whole difference lives in your head.
The AI-era plot twist
Up to this point, this is just psychology. It applies to almost any field.
But in the age of AI, there's a second layer, and this one bites harder:
The best model is genuinely, meaningfully better.
Not "slightly higher on some boring benchmark chart" better.
Better in a way that crosses an invisible threshold.
The cheaper or free models often aren't just a little worse. They can be worse in a way that actively misleads you. They quietly convince you that, "eh, AI is kind of overhyped."
It's like judging a Ferrari by sitting in a cardboard cutout at the showroom.
Technically, you saw something red.
But you did not experience the car.
A fifty-second story
Let me tell you something that happened to me.
I was an early ChatGPT subscriber — one of the first batches. I thought it was incredible from day one, so naturally I went around evangelizing it to everyone I knew.
The unpaid-intern kind of evangelizing.
Six months later, I looked up and realized some of my friends still hadn't really used it. And the ones who had were saying things like:
"Yeah, it's fine. Not as amazing as you said."
This confused me deeply.
ChatGPT is one of the most general tools imaginable. It's not some niche developer framework that three people on Earth care about. I genuinely struggled to imagine someone trying it properly and finding no use for it.
So I asked one of them:
"Show me. Walk me through exactly where it's letting you down."
He pulled out his phone.
Free version.
Previous-generation model.
Voice mode? Not even live on his account yet.
And it clicked instantly.
I didn't say a word. I just handed him my phone with my subscription open and said:
"Here, try this one."
Fifty seconds later, he hit the subscribe button himself.
Fifty. Seconds.
No lecture.
No persuasion.
No "AI will change the world" TED Talk energy.
Just fifty seconds with the real thing.
The lesson
Here's the thing.
Normally, when someone raves about a product and you say, "let me try the free version first before I commit," that's a perfectly reasonable move.
Responsible, even.
Most of the time, it's the right move.
But with AI, it can be a trap.
The free version you tried might not be the thing everyone is raving about. You think you've tried it. You may not have even found the door.
And that one missed step might not just cost you a tool.
It might cost you something genuinely big: a new way to work, a new way to learn, maybe even the entrance to a whole new era.
So spend the money.
Find your ouch number.
Think of it as buying yourself a reason to actually show up.
It's the cheapest expensive lesson you'll ever buy.