Expert Explains AI’s Impact on Entry-Level Computer Science Jobs

A One-on-One with an AI Expert

For years, the path into the tech industry was simple: earn a computer science (CS) degree, land an internship, start as a junior developer, and build a career. But AI tools write and debug basic code with ease today, shifting the market for new grads.

Dan Ye, AI expert, CEO of CollegeNode, and lecturer at UC Merced and Johns Hopkins University, explains how artificial intelligence is transforming higher education and the future of work.

TWT: With AI tools capable of writing and debugging entry-level code, the traditional pipeline of junior developers and internships is shrinking. How can recent grads build the experience employers demand when the “stepping stone” roles are being automated away?

Dan Ye (DY): Every senior engineer you admire was once a junior engineer fixing someone else’s bugs at 11 p.m. That was the apprenticeship, and AI is now very good at exactly that work.

A Stanford Digital Economy Lab study found that employment for software developers aged 22 to 25 had fallen nearly 20% from its late-2022 peak, while older developers held steady. The old deal was degree, internship, junior role, experience. The new deal is that graduates have to figure it out themselves.

The good news is that the same tools eating the junior job also turn one motivated student into a small company. With AI coding assistants, a student can ship a real product with real users in a single semester. That is the new résumé: what you built and what’s your portfolio.

Employers do not want to hear “I know Python.” They want to hear “Here is what I built, here is who uses it, here is what broke, and here is how I fixed it, here are my skills, and these are the problems I can fix.”

TWT: What does an updated Computer Science curriculum need to look like, and what specific technical skills or frameworks should CS departments start teaching tomorrow?

DY: I’ll give an example from my own class: In the investing course I teach at Johns Hopkins, AI use is not cheating. It is required, and students learn how to code agentic trading agents, prompt engineering, and keep up with the latest industry development in order to decide what stock to buy. The computer science class of the future will:

• Read and review code: AI writes; humans verify. Grade students on finding the bugs in AI-generated code, not just on producing their own.
• Testing, security and evaluation: The core skill of the next decade is knowing when the machine is wrong.
• Design systems, earlier: Architecture used to be a senior skill. Now it is the entry ticket.
• Building with AI: LLM APIs, retrieval, agents, tool calling, and the cost and latency trade-offs that come with them.
• Ship: Cloud deployment, data pipelines, and keeping something running in production.
• A second domain: Pair CS with finance, biology, law or economics. Learn to solve problems in these industries with AI. Learn how to explain AI to older, higher up management.

TWT: Parents and prospective students are increasingly questioning the return on investment for a four-year CS degree. Is the CS degree losing its absolute value, or is its value simply shifting toward different types of skills?

DY: I teach investing, so let me put it in market terms. The CS degree is a security, and it has been repriced. It has not gone to zero. Yes, New York Fed data puts unemployment for recent CS graduates at roughly 6 to 7%, above the average for all recent grads. But CS graduates still command some of the highest starting salaries of any major.

The “learn to code, get a six-figure job” premium of the last decade was a bubble of its times. Knowing how to write a for-loop used to be scarce. Now it is free. What is still scarce is judgment: knowing what to build, how the pieces fit, and when the code is lying to you.

What is losing value is the degree as a passive credential. Degree plus portfolio plus domain knowledge is still a very strong position. Degree alone doesn’t mean anything anymore unless you’re from Stanford or MIT.

TWT: Is the future software engineer someone who builds complex AI models, or someone who excels at orchestrating existing AI tools to solve business problems?

Mostly orchestrators. We will need both, but the numbers are lopsided. Frontier models will be built by a small number of people at a small number of companies, much like advanced chips are made in a handful of fabs.

Most software engineers will look more like general contractors. They will know which tools exist, how to combine them, where each one fails, and how to translate a messy business problem into a working system. The future engineer is part architect, part product manager, part auditor, and part problem solver.

TWT: Companies are rapidly adopting AI to reduce headcounts for basic technical tasks. Are corporations lowering their intake of junior talent too quickly, and what long-term talent shortage risks are they creating by eliminating entry-level onboarding?

Many companies are eating their seed corn. Cutting junior hiring looks great on this quarter’s margins and terrible on the ten-year chart.

Senior engineers do not grow on trees. They are junior engineers with ten years of mistakes behind them. The Stanford researchers note that AI mostly replaces codified, textbook knowledge, while experienced workers hold tacit knowledge: the instincts and context that come only from doing the job. Eliminate the junior job and you eliminate the factory that produces seniors.

Junior engineers also tend to be those who come with new ideas. Eliminating them runs the danger of the company going into an echo chamber, where only old ideas survive and live. The bill comes due in five to ten years. Expect a shortage of senior talent, a spike in senior wages, and companies running AI systems that nobody inside the building fully understands. That last one should worry every board.

The smart companies will still hire juniors, perhaps fewer, but train them differently. Make them the reviewers, testers and auditors of AI output, and pair them with seniors. Build an AI-era apprenticeship rather than abandoning apprenticeship altogether. My advice to students is the same advice I give investors: don’t compete with the machines. Learn to own them.