Socrates Didn't Sell Courses

There are roughly 47,000 courses on cloud architecture available online right now. Udemy, Coursera, A Cloud Guru, LinkedIn Learning, YouTube playlists with thumbnails where someone points at a diagram. Forty-seven thousand.

Go ahead and search. I’ll wait.

Now here’s a question. If 47,000 courses exist on cloud architecture, and cloud misconfigurations are still the number one cause of security incidents, and companies are still routinely surprised by their AWS bills, and engineers are still deploying Kubernetes clusters they don’t need because “that’s what you do”… then what exactly are 47,000 courses teaching?

Hold that number. We’ll need it later.

What are you actually paying for?

Knowledge has never been more available. This is not a controversial statement. It is plainly, observably true. You can learn Terraform for free. You can watch a senior engineer live-code a CI/CD pipeline on Twitch. You can read the Kubernetes documentation (good luck, but you can). The supply of technical knowledge is essentially infinite.

So why are the same mistakes happening at the same rate?

If knowledge were the bottleneck, 47,000 courses would have fixed it by now. So what is?

A course on cloud architecture will teach you how to build a multi-tenant SaaS. It will walk you through VPCs, subnets, security groups, load balancers. By module 12, you will feel confident. You will feel ready. You will have a certificate.

But will the course ask you whether you should build a multi-tenant SaaS in the first place? Whether your two paying customers actually need tenant isolation? Whether the complexity you’re about to introduce will outlive the problem it solves?

Where, in 12 modules, does that question fit?

The person who did everything right

Last year I sat across from an engineer who had done everything right. Three AWS certifications. A Solutions Architect Professional among them. He’d followed best practices from two different courses, built exactly the architecture the instructors recommended, and deployed it to production for a platform with eleven users.

Eleven.

He had a multi-AZ setup with auto-scaling groups, a managed Kubernetes cluster, a service mesh, and a CI/CD pipeline that would make a Fortune 500 proud. His AWS bill was $4,200 a month. For eleven users. When I asked why he’d built it this way, he looked at me like I’d asked why he breathes. “This is the reference architecture. This is what the course taught.”

He knew how to build everything and when to build nothing. Three certifications, and not one of them had included a module called “Is This Worth Building?”

Was he incompetent? That’s the comfortable answer, isn’t it? But what if the problem is the opposite? What if he was so competent at executing that nobody, at any point in his education, had asked him to stop and think about whether he should?

How do you certify someone in knowing when not to act? What does that exam look like?

The incentive you’re not supposed to notice

So you might think the problem is bad courses. It isn’t. Let me make it worse.

Imagine you’ve built a course on cloud architecture. It’s good. It sells. People learn things. You’re genuinely helping.

Now one of your students asks: “I’m spending $4,000 a month on AWS and I don’t know why. My architecture follows the best practices from your course, but the bill doesn’t make sense.”

This is a FinOps problem. Your course doesn’t cover FinOps. So what do you do?

If you’re honest, you say: “That’s outside the scope of this course. You need to understand cloud financial governance, cost attribution, reserved instance strategy, and probably rethink some of the architectural choices I just taught you.”

But if you’re a business, you say: “Great news! I have a FinOps course launching next month.”

Do you see what happened? The answer to “I followed your advice and something went wrong” became “buy more of my advice.” Not because the course creator is dishonest. They might be brilliant and well-intentioned. But the incentive structure of selling packaged knowledge requires that every question leads to another package. Doubt is bad for conversion rates. Certainty is the product.

But here’s the thing. Even if you fix the incentive, even if the course creator is a saint with no business model at all… does the problem go away? Can you put “the ability to sit with uncertainty” in a slide deck? Can you grade someone on “resisting the urge to reach for the right answer before understanding the question”?

If the answer is no even without the incentive problem… then what were we really talking about?

Am I just being elitist?

I can already hear the objection, and it’s a good one.

“Courses help people. Real people, who can’t afford mentors or university or your Socratic wine-bar philosophy. A $29 course on AWS fundamentals is genuinely transformative for a developer in Lagos or Dhaka who has a laptop and an internet connection and zero access to the old-boys network of tech conferences and consulting connections. Are you really sitting here, arguing that courses are bad?”

No. I’m not. That objection is correct.

Courses genuinely democratize technical knowledge. A 20-minute video reaches thousands of people. A Socratic dialogue reaches one person at a time. If I had to choose between a world with 47,000 courses and a world with none, I’d take the 47,000 without hesitation.

So sit in that for a moment. If courses are good (they are), and they help people (they do), and they make knowledge accessible at scale (undeniably)… then what exactly am I questioning?

What if the thing that’s missing can’t be scaled? What if you can’t democratize the one skill that makes all the other skills useful? And if you believe knowledge should be accessible to everyone… what do you do with the thing that refuses to be packaged?

So why aren’t you selling a course?

I should probably mention that I could. Build courses, I mean. I’m involved in multiple companies. I manage roughly 45 projects across consulting, products, and open source. I’ve sat across from that engineer with the $4,200 bill, and from dozens like him. Any of these stories is a Udemy bestseller waiting to happen.

I chose not to. Not because courses are wrong. Because nobody would buy this. What would the landing page say? “Pay $49 to become less certain about everything”? “By the end of this course, you will have more questions than when you started”? Would you buy that? Would you rate it five stars?

So who’s the customer? You need critical thinking to know you need critical thinking. (Do we love recursion?) The people who would benefit most from questioning their certainty are, by definition, certain they don’t need to. What does a market look like when the buyer has to already own the product to want it?

The question behind the question

There’s a concept in ancient Greek philosophy called aporia. It translates, roughly, as “a productive impasse.” It’s the moment in a Socratic dialogue where you realize that the thing you were certain about five minutes ago now has a crack in it. Not because someone proved you wrong. Because someone asked the right question, and the certainty couldn’t survive the asking.

Let’s go back to that number. 47,000 courses. How many of them include a module called “When Not to Use This”? How many course creators have a financial incentive to tell you that you don’t need their course?

But here’s what I keep circling back to. You just read 2,000 words questioning the value of packaged knowledge. You probably nodded along. You might even share it. You might feel like you’ve learned something.

Have you, though? Or did you just complete Module 1?

If this essay made you feel smarter, more certain, more equipped to see through the course-selling machine… then I’ve done exactly what I criticized. I’ve packaged a feeling of knowing and handed it to you. The only difference between this and a Udemy course is that I didn’t charge you.

So. Did reading this teach you to think? Or did it just teach you what to think about thinking?

I can’t answer that for you. And if I could, you shouldn’t trust the answer.