Nobody Wants to Learn AI

⭐️ (Not a Medium member yet? Read the rest of this story paywall-free here.)

Last week, someone in your feed posted about being "thrilled to build AI agents this weekend." Forty-seven likes. Fire emojis. "Love the growth mindset!" You stared at it and felt something you couldn't name. Not jealousy. Not inspiration. Something closer to recognition. The way you recognize a mask because you're wearing the same one.

Nobody is thrilled to learn agent frameworks on a Saturday. They're afraid of what happens if they don't.

You probably have your own version of this. A browser tab you keep meaning to open. A course you bookmarked during a sale and never started. A Slack thread about "AI readiness" that you skimmed and closed. The quiet admission that you don't know enough to stay relevant, and the quieter admission that bookmarking the resource let you feel like you were doing something about it.

You've been writing production code for years. By most reasonable measures, you know what you're doing.

And yet.

That tab sits there. Glowing faintly. A talisman against obsolescence.

The Same Activity, Two Different Meanings

Early in your career, you learned a framework because it was genuinely exciting. A new mental model for building interfaces. You stayed up late not because you were anxious but because you couldn't stop. The learning felt like building. Like adding rooms to a house you were just beginning to inhabit.

Now you open a course tab because your company started an "AI readiness" initiative. There's no mandate, technically. Just a Slack message from leadership about "staying ahead of the curve" and a shared spreadsheet where you can log your upskilling hours. Voluntary, of course. The way salary negotiations are voluntary.

The distinction matters because the industry doesn't acknowledge it. Every conference talk, every LinkedIn post, every corporate learning initiative treats all learning as the same species. Growth. Development. Curiosity. Whether you're a new grad exploring your first language or a career-changer using Coursera to escape a dead-end job or a 15-year veteran trying to prove you're not a dinosaur, it's all filed under the same aspirational banner.

But the body knows the difference. Curiosity feels like leaning forward. Defense feels like bracing.

I want to be clear about something: for some people, these platforms are genuinely life-changing. The self-taught developer who used a Coursera scholarship to move from data entry to engineering, the career-changer who learned Python on free YouTube tutorials. That's real. That matters. This isn't about those people. This is about a system that takes their stories and uses them as marketing for something very different: the perpetual anxiety machine that tells experienced professionals their decade of work might expire next quarter.

The Thing You Bought vs. The Thing You Got

Here's where the economics get clarifying.

The average MOOC completion rate is below 10% for free-track learners, with some studies putting it as low as 3%. More than nine out of ten people who enroll in an online course never finish it. They buy access. They don't buy knowledge. They buy the feeling of progress, the same way a gym membership on January 2nd buys the feeling of fitness.

The corporate side isn't better. A 2014 study by Gartner found that 45% of corporate training qualifies as "scrap learning," content employees complete but never apply on the job. That Kubernetes course you watched on double-speed during lunch breaks? By performance review season, you'll have retained the certificate and almost none of the knowledge.

The upskilling industry doesn't sell knowledge. It sells the feeling of doing something, and even that feeling has a completion rate of 3%.

The Skill That Expires vs. The Skill That Compounds

This is the part nobody talks about, because talking about it would break the business model.

Not all skills depreciate at the same rate. Your knowledge of a specific framework has a half-life of about two to three years. React hooks, Kubernetes configurations, the agent framework API you learned last month: the specifics you picked up this quarter will be partially obsolete by next year and largely irrelevant in three. This isn't a failure of learning. It's the nature of implementation knowledge. It's perishable by design.

Your ability to debug a system you've never seen before has a half-life measured in decades. Architectural judgment, the instinct for where complexity will hurt you later, the capacity to read a codebase and understand not just what it does but why someone built it that way: these are durable skills. They compound with experience instead of depreciating with time.

Think about the SRE on your team who has spent eight years keeping a payments system alive through migrations, acquisitions, and near-catastrophic failovers. Who knows where every brittleness hides. Who can diagnose cascading failures from the shape of a latency graph the way a cardiologist reads an EKG. Now imagine their manager asking what they're doing to "stay current with AI." Their entire career is a durable skill. The system that evaluates them can't see it.

The upskilling industry has no product for that kind of expertise. You can't package it into a six-week course with a certificate. It develops slowly, through years of building and maintaining real systems, through the painful process of watching your own architectural decisions age and learning which ones held. There is no shortcut. There is no 90%-off sale.

So the industry sells you the perishable kind. Over and over. React today, agent frameworks tomorrow, whatever tool emerges next quarter. Each course addresses a skill with a two-year shelf life, which means you'll need another course in two years. And another after that. The business model depends on your knowledge expiring.

Your React knowledge has a half-life of about two years. Your ability to debug a system you've never seen has a half-life of decades. The industry charges you quarterly for the first and has no idea how to price the second.

I should say the obvious thing: sometimes you do need the perishable skill. Sometimes the new framework is the right tool and learning it is the right call. And if you're four years in, still building your foundation, the cruelest part is this: you're being told the perishable stuff is all that matters before you've had time to develop anything durable. The system is pricing your potential at zero while demanding you pay full rate for currency.

The problem isn't learning new things. The problem is an economy that treats perishable and durable skills as interchangeable, prices them identically at hiring time, and then acts surprised when experienced engineers feel like they're running in place. Durable skills make you a terrible customer. Perishable skills make you a subscription.

The Confession I Keep Editing

I should be honest about something, and more honest than I usually am.

I build AI systems. I know what these tools can and can't do. And I still feel it. The pressure to signal that I'm keeping up. The impulse to post about the latest framework on LinkedIn not because I learned something meaningful, but because the keyword matters more than the knowledge. I know it, and I've done it anyway.

I've watched engineers I respect get evaluated on whether they "adopted AI tools" instead of whether they built systems that worked. I've seen performance rubrics that reward course completion and penalize the kind of deep, quiet expertise that keeps production systems alive. The rubric measures compliance, not capability. It rewards the perishable and ignores the durable. And the people filling out those rubrics know it. They tighten their jaws and check the boxes because the alternative is admitting the system they're enforcing is broken.

I know because I've been on both sides of that table. The feeling is the same: you're performing competence instead of practicing it.

The One Question That Tells You Everything

Next time you're about to open a course, a tutorial, a weekend project with a new framework, ask yourself one question:

Will I use this in the next six months, or am I learning it because having it on my profile makes me feel safer?

The first is investment. The second is a minimum payment on a debt that keeps growing.

If the answer is "I'll use it," learn it. Learn it deeply. Build something real with it. That's not anxiety. That's craft.

If the answer is "it makes me feel safer," close the tab. Spend that hour reading the codebase you already maintain. Understand the system you already own. Build the durable skill that no course can teach and no certificate can prove. The thing that makes you the person in the room who says "this will break in six months" and is right.

The upskilling industry can't sell you that. Which is exactly why it's valuable.

Competence Debt

That tab is still open.

Some mornings I open my browser and it's there, between Jira and Slack, glowing with the particular patience of things that know they've already won. I'll click it eventually. Not because the course will teach me something durable. Not because it will make me meaningfully better at the work that actually matters. But because the credential economy demands proof of currency, and currency is what expires.

Somewhere, someone is genuinely excited about AI agents, staying up late because the possibilities feel electric. Not because of a performance review. Not because of a Slack message from leadership. Because the thing itself is interesting. That feeling still exists. Most of us just can't reach it through the anxiety anymore.

Here's what I've started calling it: competence debt. The accumulation of perishable certifications while your durable skills atrophy from neglect. Every hour spent on a course that teaches you the API of the moment is an hour not spent understanding the system you've been maintaining for three years. Every certificate is a minimum payment on a debt that keeps growing. You feel productive. Your profile looks current. And underneath, the skills that would actually make you irreplaceable are quietly compounding interest in the wrong direction.

The industry taught us to call this growth.

The rubric was designed to measure it.

The word we were all looking for is depreciation.