Stop Telling Students Computer Science Is Dying (opinion)

Every computer science department head I know is having a version of the same conversation—with a parent at orientation, with a high school senior at an open house, with the colleague at lunch who asks if it’s true. The headlines say computer science is collapsing. The data shows something else. Failing to make that clear to students and families leaves them with no choice but to decide based on the wrong information.

Surrounded by Literalists

A colleague of mine—not a computer scientist, but someone who works alongside them every day—laughed at me recently after I took something they said a little too literally.

“I’m surrounded by literalists,” they said.

They meant it as a joke. I received it as a compliment. It was not the first time I’ve had some version of this exchange.

The literalism they described is not a personality quirk. It is a trained habit of mind—one that computer science education cultivates deliberately. A computer does only and exactly what you tell it to do. Not approximately. Not what you meant. What you said. Computer scientists learn to translate underspecified requirements into unambiguous instructions, to anticipate every edge case. We develop, over years of practice, an instinct for exactness.

I love writing code. So do most of my colleagues. And the particular discipline we love—sitting with a blank screen and producing working code, character by character, from our own reasoning—is becoming less central as AI coding assistants get better. The reasoning is still ours; the classic coding increasingly is not. Pretending otherwise would be naïve.

But learning computer science is less about the code you produce and more about the habits of mind you build. Writing code from scratch teaches you to think in systems, to reason about cause and effect, to hold an entire logical structure in your head and interrogate it. Those capacities do not become obsolete when the tools change.

The Headlines Versus the Data

If you have been paying attention to recent technology news, you have encountered headlines like “Goodbye, $165,000 Tech Jobs” and “Graduates Reset Ambitions in Pursuit of First Jobs.” The message is unmistakable, and for the student considering a computer science major, it is genuinely frightening.

So I did what a CS education taught me to do automatically: I looked at the data, specifically data from the U.S. Bureau of Labor Statistics’ Current Employment Statistics. And I did the following exercise with my first-year CS majors at Virginia Tech on the last day of the semester, starting with a chart I had recreated from a widely circulated visualization—one that had been shared with me as justification for discouraging students from studying CS. No title, no explanation, just the chart. “What do you see?” I asked them.

They saw what most people see: The trend since 2022 heads sharply downward. Then I asked how it made them feel. Hopeless, said one. Despair, offered another.

A beat of silence. Then a hand went up. “It looks like the industry was way overhiring around 2022.”

Right observation, but that cliff still loomed.

I asked the students to look more carefully at the chart. The y-axis shows the change in jobs compared to the same month one year earlier—a choice that virtually guarantees a dramatic visual when growth slows after a boom. It is not lying, exactly. But it is designed to alarm rather than inform.

I then showed my students the same data plotted differently—as total employment rather than year-on-year change.

The cliff disappears. What you see instead is a line that grew from roughly 900,000 jobs in 1990 to a peak of just over four million in 2023—not 2022, as the first chart might suggest. The peak on that first chart is an illusion created by the choice of what to measure.

But the story doesn’t end there. The dataset behind both charts counts jobs at tech-sector employers.

A second Bureau of Labor Statistics dataset asks not where people work, but what they do. It surveys employers across every sector of the economy and counts how many workers are doing computer and mathematical work, regardless of what industry their employer is in. A software developer at a hospital, a data scientist at a bank, a network engineer at a car manufacturer—all of them appear in this dataset, but not in the first one. A receptionist, janitor or human resources professional at a tech company? They’re in the first dataset but not the second. I showed my students another chart—just the line, no numbers on the y-axis—and asked them a simple question: Do you think this line sits above or below the black line in the previous chart?

Most guessed below (though by this point in the lecture, some of them were on to me). The assumption was intuitive: Surely most CS jobs are at tech companies. Their guess turned out to be wrong, and the gap between what people assume and what the data shows is exactly the kind of thing a CS education trains you to find.

Computer and mathematical occupations employed just over two million people in 1997. By 2024, the number was 5.2 million, and it has barely flickered during the recent tech layoffs.

Why the divergence? Because the layoffs happened at tech-sector employers, but the workers themselves did not disappear. They moved. Into finance. Into health care. Into government. Into every sector of the economy that runs on software, which is to say every sector of the economy.

The Supply Gap No One Is Talking About

My students absorbed the job numbers, but I could see the next worry forming—the fear that a market with millions of jobs also has millions of graduates competing for them. I added in a data point from one more federal dataset: the number of working-age people in the labor force who hold a computer science degree. In 2024, that number was approximately 2.8 million.

The job count was 5.2 million.

That gap—CS jobs filled by people without CS degrees—is persistent. The field has never been credential-constrained. Its need for trained people has always outstripped the supply of CS degree holders.

This does not mean the market has not shifted. It has. The entry-level role that once consisted primarily of writing straightforward code under close supervision is genuinely changing, in part because AI tools are increasingly good at exactly that task.

But that shift demands people who can work with AI to complete these tasks, who can decompose a problem before any code is written, who can debug not just a program but a dataset, a model, a set of assumptions. That kind of thinking—the productive literalism of CS training—transfers across every domain technology touches. It is the genuine analytical foundation, the mark of the person who understands the problem deeply enough to know whether the AI’s answer is right.

What CS Actually Teaches

As we neared the end of class, we took a break from the data, and I asked my students a different kind of question. If AI is even more capable two years from now (and it will be), if entry-level coding is automated and teams are smaller (and they will be), what should an ideal CS graduate focus on?

They did not say, “Learn more programming languages.” They said, “Understand fundamentals. Think at the level of systems. Know how to debug. Think critically. Collaborate.” But even more telling were the things they suggested to stop overinvesting in. Stop obsessing over detailed syntax and mechanics. Stop overrelying on AI. And cultivate skepticism—toward AI outputs, data, headlines and even your own first instincts.

These are first-year students, one semester into a CS degree. They had already intuited a central tension of our field: that the tools will keep changing, so the thinking underneath the tools is what matters most. That thinking has a name—two names, actually: skepticism and discernment. Skepticism is the posture: You don’t accept the output just because it’s fluent and confident. Discernment is the harder skill: knowing, once you’ve doubted, which part is right, which part is wrong and why. Every CS student learns these through testing and debugging—deciding to write a test is skepticism; tracing why a result failed is discernment.

You cannot ship vague code. You cannot approximate a data structure. You cannot debug by vibe alone. The field is, at its core, a training in literalism, and that rigor changes not just how you think about computer programs, but about everything.

A Field in Radically Exciting Change

Computer science is going through a genuinely transformative moment. But transformation is not the same as collapse. The 2022–23 tech-cession—real and painful for the people who lived it—appears, from the occupations data, to have been a tech-sector employer phenomenon, not a CS worker phenomenon. The work simply moved somewhere else, and that somewhere else is now everywhere.

The millions of CS workers already embedded outside dedicated tech firms represent the leading edge of a pattern that will only continue. The question for a prospective student is not whether CS skills will be valuable; the data is unambiguous on that. The question is what kind of CS training prepares you for a world where the tools keep changing.

Here is what I told my students.

Use AI as a tool, not a crutch. Get genuinely good at something—not everything, but something. Learn how to work with others: explain your thinking, ask good questions, communicate clearly and often. Build things you can show people. Be the person who deeply understands the problem being solved and not just the tools you’re using to solve it. And pay attention to where tech meets something else.

Consider the CS-trained professional leading data-informed initiatives to address homelessness—building systems to track, analyze and act on the complex, evolving patterns of a population in crisis. Or the one navigating the intersection of public policy, health-care IT and mental health information exchange, translating between the technical realities of data systems and the policy decisions that determine how care gets delivered. Neither of them is writing code every day. Both are doing work that only becomes possible when CS meets something else.

If you skim the headlines, you might think the options of a CS graduate are shrinking. If you read them the way a CS education teaches you to read everything—with skepticism, and then discernment—you might instead see that they are simply changing and that computer science is precisely the discipline that teaches you to tell the difference.