Learn to Code, They Said: How the Best Career Advice of 2020 Became 2026's Cruelest Joke

There’s a scene in The Graduate — the 1967 one with Dustin Hoffman — where a middle-aged family friend corners young Benjamin at his graduation party. He leans in close, dead serious, and whispers one word of career advice: “Plastics.”

It’s supposed to be the golden ticket. The future. The sure thing.

Benjamin just stares, confused. The audience laughs because they already know how the story ends: the sure thing was never sure.

For our generation, that word wasn’t plastics. It was code.

The Promise

Between 2015 and 2021, “learn to code” became the universal answer to every career question. Laid off from a factory? Learn to code. Stuck in retail? Learn to code. Journalism degree collecting dust? Learn. To. Code.

Presidents said it. Bootcamps raised billions on it. LinkedIn influencers built empires selling the dream. Every tech recruiter from San Francisco to Bangalore repeated the same pitch: the world needs developers, and it always will.

So people listened. They quit jobs. They took out loans. They spent nights on freeCodeCamp and weekends grinding LeetCode. They got CS degrees, bootcamp certificates, and GitHub profiles full of green squares.

They did exactly what they were told.

The Punchline

The American Prospect published a piece three days ago titled — I’m not making this up — “Learn to Code, They Said.” The subtitle should have come with a trigger warning for anyone who graduated between 2022 and 2025.

Stanford’s Digital Economy Lab found that early-career workers aged 22 to 25 in AI-exposed occupations like software engineering experienced a 16% relative employment decline. Not a slowdown. A decline. While employment in less AI-exposed jobs stayed flat or grew.

Meanwhile, the broader numbers are brutal. According to tracking data, the tech industry laid off nearly 80,000 employees in just Q1 2026 — and almost half of those cuts were explicitly attributed to AI. The year-to-date total? Over 143,000 people across 342 companies. That’s roughly a thousand people per day losing a job they were promised would be recession-proof.

We wrote about this back in April — companies aren’t even firing people because AI is doing the work. They’re firing people because a PowerPoint deck says AI might do it someday. The firings are based on vibes. Which, given the era, feels appropriate.

The Block Treatment

If you want a case study in how “learn to code” collapses in real time, look at Block.

Jack Dorsey’s company — the one that makes Square and Cash App — cut 40% of its workforce. Over 4,000 people. Gone. Then Dorsey made every remaining employee use AI tools daily, with AI fluency baked directly into performance reviews. Not optional. Not “try it out.” Use it or your review suffers.

One employee told reporters that adoption was “not optional” and that layoffs “became the enforcement mechanism.”

Think about that for a second. You learned to code. You got hired at a respected fintech company. You were good at your job. Then your CEO decided that coding the old way is over, fired half your friends, and told you that your review now depends on how well you collaborate with a chatbot.

The employee quote that stuck with me: “If the tool were good, we’d all just use it.”

That’s the quiet part nobody’s saying out loud. If these AI tools were genuinely 10x better, you wouldn’t need mandates. You wouldn’t need to tie adoption to performance reviews. People adopt things that make their lives easier. They resist things that make their work worse while their employer pretends otherwise.

The Skills Paradox

Here’s where it gets truly absurd.

A TechSpot report found that programmers forced to use AI coding tools at work say their skills are deteriorating. They’re not writing code anymore — they’re reviewing AI output. They’re becoming editors of machine-generated slop, and the muscle memory that made them valuable is atrophying.

We explored this in depth with the story of a developer who forgot how to code after four months of pure AI prompting. She wasn’t lazy. She was doing what her employer asked. And her brain quietly stopped retaining the patterns she’d spent years building.

So the cycle is:

Learn to code (as advised)
Get a coding job (as promised)
Company mandates AI tools (not as planned)
Stop actually coding (because the AI does it now)
Lose the skills that got you hired
Get laid off because “AI handles this now”
Try to find a new job, but your skills have atrophied
Somebody suggests you “learn to code” again

It’s not a career path. It’s a hamster wheel designed by a management consultant who read one blog post about GPT-4.

So Is Coding Dead?

No. But “learn to code” as career advice is dead, at least in the way it was originally meant.

The advice assumed that writing syntax was the hard part and that demand for people who can type for loops would grow forever. It didn’t account for AI becoming decent at exactly that — producing functional code from natural language prompts.

What AI still can’t do: understand why. Why does this mobile app need to handle offline state differently on a subway versus an airplane? Why does this payment flow need to retry differently in Australia than in Germany? Why does this SwiftUI view need a custom layout when the standard LazyVGrid technically works but feels wrong to users?

That “why” layer is where real developers live. It’s architecture. It’s product thinking. It’s knowing that the code compiles but the user experience is garbage. No AI mandate at Block is going to automate that.

If you’re learning iOS development, for example, the value isn’t in memorizing UITableView delegate methods. It’s in understanding when SwiftUI’s declarative model breaks down, how to structure an app so it survives three years of scope changes, and how Apple’s frameworks actually behave at the edges. That’s the knowledge that our SwiftUI courses focus on — not syntax, but the judgment layer that sits above it.

The New Career Advice

If someone corners you at a party in 2026 and whispers career advice in your ear, here’s what they should say:

“Think.”

Not “learn to code.” Not “learn to prompt.” Think. Understand systems. Understand people. Understand why software exists and who it serves. The tool you use to write the code — whether it’s Xcode, Cursor, or an AI agent — matters less every year. The ability to decide what to build and why? That matters more every year.

The generation that followed “learn to code” advice isn’t stupid. They worked hard. They did the right thing according to every signal the market gave them. The market lied, and the people who gave that advice aren’t the ones paying the price.

If you’re reading this and you’re one of those people — the ones who learned to code because you were told it was safe — I’m sorry. The advice was given in good faith by people who didn’t see this coming. The companies that hired you and then mandated their way out of your job? They don’t get that excuse.

The skills you built aren’t worthless. They’re incomplete. The code was always just the medium. The thinking was always the point. Nobody told you that part, because it’s harder to sell a bootcamp called “Learn to Think.”

But it’s the only advice that ages well.