The End of the ‘Learn to Code’ Era

Over the past few months, Meta, Coinbase and Block have each laid off at least 10 percent of their workforce—roughly 13,000 jobs combined—and each pointed, at least partly, to artificial intelligence as the reason. They are not alone. More than 200 tech companies have cut an estimated 124,000 jobs so far this year.

For students sitting in our classrooms, the message lands like a verdict: The path we sold them is closing. On the first day of classes last fall, a freshman—call him Alex—introduced himself and explained that he had already changed his major. He had arrived intending to study computer science, but over the summer he’d heard so much from relatives and friends about layoffs and the weakening tech-job market that he began to worry about betting everything on it.

Before classes even began, he had switched to computer engineering—close enough to use his existing computer skills, but different enough to feel safer—and added an artificial intelligence minor as a further hedge.

Alex is right that learning to code is no longer a key to a guaranteed future. But he’s dead wrong that computer science is dying. We sold him a bill of goods, and the bill is coming due.

For nearly two decades, a bipartisan consensus about a surefire path to good jobs dominated both American education policy and dinner-table conversations: Learn to code. This premise was not just rhetoric but was enshrined in legislation and job-training initiatives throughout the country. It was a quick fix for social mobility. We told laid-off coal miners, struggling journalists and working-class kids that syntax was salvation, higher education’s guaranteed return on investment—a direct pathway to the middle class.

That certainty is now fracturing.

Undergraduate enrollment in computer science at four-year U.S. colleges fell by an average of 8.1 percent in fall 2025, per the National Student Clearinghouse, and graduate enrollment in computer science programs fell by 14 percent. The entry-level tech market has tightened; tech job postings on Indeed were down 36 percent from their early 2020 levels as of July 2025. AI is killing coding as a profession, so people think computer science must be dying. The machines can now lay pipe, and the human would-be plumbers are panicking.

This panic is the result of a narrative we helped create. Big tech wanted more software engineers, CS departments wanted more students and policymakers wanted an answer to job losses in other sectors. So we built a simple story: Coding is the answer. It never was, and AI is now exposing that.

The spine of our programs—through all the theory and methods courses— was programming. After all, computer programs are the objects we design, analyze and build. And coding let us promise workforce readiness to our funders and career prospects to our graduates.

But that focus on the practical came at a cost. Even as computing spread everywhere, we kept teaching it as if it was mostly about the machine. We underweighted the human elements of technology and mislabeled communication, teamwork, user-centered design and ethics as “soft skills,” treating them as optional. In reality, they are among the hardest and most essential competencies in computing. They distinguish systems that merely function from systems that people can trust. You can automate information processing, but you cannot automate the moral weight of systems that deny loans, accuse people of crimes or rerank medical waiting lists. Understanding how such systems behave, fail and are governed isn’t an elective. It’s exactly what computer science is for.

And yet we taught students that the field was coding. When machines learned to do that, the discipline appeared to dissolve.

But programming is just a tool—it is to computer science what calculation is to mathematics, or microscopy to biology. Computer science is concerned not with code per se, but with the design, behavior and failure modes of complex information systems, both technical and human.

The irony of the AI era is that as AI reduces the importance of writing code, the more vital the deeper habits of mind fostered by computer science become.

The management of complex information systems is much larger than just software—the question of how to deal with AI automation embedded in human information systems is critical. Designing and debugging the complex systems of AI embedded in software and in human institutions is among the most fundamental disciplines for the new era. Indeed, computer science should now be considered a cross-cutting foundational discipline, similar to mathematics or writing, providing core concepts and methods applied nearly everywhere.

We must also reckon with the social wreckage that could follow a retreat from computing education. However imperfectly, “learn to code” reflected a real democratic impulse: technical literacy as a pathway into one of the most economically and culturally powerful sectors of modern life.

When automation compresses the lower rungs of a skilled field, it does not distribute opportunity at the upper rungs more widely—it concentrates it. Document-review software did not democratize law so much as consolidate it; accounting software narrowed the market for bookkeepers even as top CPAs remained in high demand. There is no reason to expect the software industry to behave differently as long as coding is the focus. The entry-level roles that once brought first-generation graduates and community college transfers into the industry are among the first being compressed or displaced.

If institutions respond by shrinking CS education, they are not just neutrally adapting to market signals. They are choosing, by default, who gets to shape the systems that govern modern life—who builds the hiring algorithms, who designs the predictive policing tools, who decides how a medical triage system ranks patients. That technical class is already small and homogeneous. Pulling back makes it smaller still. Technical power doesn’t disappear. It concentrates and compounds.

Generative AI hasn’t killed computer science; it has revealed the limits of a story that equated the discipline with its most basic tool. Alex’s instinct wasn’t wrong—if computer science is just coding, its days are numbered. But it’s not. What Alex needs, and what we failed to offer him, is the real thing: the study of complex systems, how they are designed, how they fail and how they are governed. His AI minor, added as a hedge, points in the right direction—computing embedded in something larger, applied to something that matters. He found the right answer for the wrong reasons.

That’s not his failure; it’s ours. Now that AI has revealed the hollowness of the “learn to code” promise, the need for true computer science is greater than ever. The question is whether our institutions will have the courage to reframe our teaching and stewardship of the field accordingly. If we continue to relate to computer science as a narrow labor pipeline, we will watch access contract and control over tech become more concentrated and less governable.

Or, we can see computer science properly as the foundational discipline for a world organized by software and AI. Someone still has to frame problems, assess outputs, understand complex systems and ensure accountability when they fail. In the age of widespread AI, this is indispensable. The only question is whether we have the resolve to rethink our programs accordingly.