Pipeline
The CS Degree Question Is a First-Job Market Question
A computer science degree's forty-year value and a graduating class's twelve-month problem are two different questions, and most answers pick one and pretend it settles the other. The honest read separates them. The first-job market for new graduates has genuinely tightened, part of it for reasons that predate generative AI entirely, while the measurements that exist find full replacement of a work skill rare and none break out programming on its own. What changed inside the work is what should change how you spend the degree.
The takeNeither side of this argument is being straight with a student weighing four years and a large bill. A university answering with graduate salary averages is answering about the degree's lifetime value while a reader is asking about next June. Doom content answering with a screenshot of a model writing a function is answering about programming as an activity while a reader is asking whether anyone will hire them to do it. Both dodges are comfortable because neither requires citing a number that could be wrong. The first-job data is messier and more useful than either.
Where Olive fits
Open a role and see what the work shows
Where Olive shows up in this same pipeline, it never stands for a degree or its absence: six evidenced findings about one candidate's work on a real assignment, written by a person, and the candidate reads the identical report the employer does, free.
Rank your shortlistName the Question Both Sides Dodge
A university's answer to "is CS still worth it" is almost always about the degree: lifetime earnings, alumni outcomes, the return on tuition averaged across a career. A viral post's answer is almost always about programming as a skill: a screenshot of a model writing working code, framed as evidence the discipline is ending.
Both are real questions, and neither is the one a graduating senior actually has, which is narrower and more urgent: will I get hired in the next twelve months, and how much of the difficulty is AI versus something else entirely. That narrower question has an evidence base of its own, and using it does not require picking a side in the degree-versus-doom argument first. It requires looking at what has actually happened to new graduates lately, separately from what a model can currently do with a coding prompt, and holding the two apart on purpose.
What the First-Job Data Actually Shows
The first-job evidence that can be verified is broader than computer science and useful anyway. The Burning Glass Institute found that 52% of the US Class of 2023 were, one year after graduation, working in jobs that did not require a degree at all, spread across the range of majors, with more than a quarter of engineering graduates among them 1. Underemployment there is Burning Glass's own composite across all majors, not a computing figure and not a government statistic.
The precise recent-graduate unemployment rate broken out by major, the number this question really wants, is absent on purpose. The source that publishes that breakdown could not be opened to check it, so no figure from it appears here rather than being repeated secondhand. Treat any precise major-specific number you meet elsewhere the same way: ask which source and which release date it came from before letting it move a decision.
As a neutral baseline for sizing any youth-unemployment headline you encounter, the US unemployment rate for people aged 16 to 24 ran roughly double the national rate through mid-2026, 8.5% against 4.1% in July 2. That series mixes teenagers still in school with graduates, and a gap of about two to one is what a labor market with people finishing school and starting out ordinarily produces, so the number carries no information about AI by itself. Neither figure tells you what happens to a computer science graduate specifically. Both tell you the ground a computer science graduate is standing on, which is the honest limit of what this evidence can do.
Don't Blame AI for All of It
The Burning Glass finding comes with a sentence worth repeating exactly: the underemployment trend predates both the pandemic and the rise of generative AI, which points to structural forces in graduate supply and entry-level demand that were already at work before ChatGPT existed 1. A hiring correction that started for reasons unrelated to AI is easy to misattribute once a new technology arrives loudly in the same window.
Nor is the entry-level squeeze a technology-sector story. Stanford's Digital Economy Lab, working from ADP payroll records, reports that employment of 22-to-25-year-olds in the most AI-exposed occupations has fallen behind their less-exposed peers since 2022, through fewer people being hired rather than more being let go, and the authors call these descriptive indicators rather than causal estimates of AI 3. Their occupation-level estimate is essentially unchanged when computer occupations are dropped from the sample and only slightly smaller when technology firms are dropped, which is their own evidence that the findings are not specific to technology roles 3. If your worry is that computer science was singled out, that is the part the strongest available evidence speaks to, and it does not support the read.
Where Did the First Rung of Coding Actually Move?
Programming as an activity has not been replaced wholesale, even where the fear is loudest. Indeed's Hiring Lab rated almost 2,900 work skills found in US job postings against what current models can do and judged 19 of them, 0.7%, very likely to be fully replaced, up from zero in the previous edition 4. Those ratings are models scoring skills rather than observations of work being done, and a skill is not a job.
The same index splits everything it rated four ways: 40% of skills minimal transformation, 19% assisted, 40% hybrid and 1% full 5. Hybrid is its label for AI changing how the work gets done while human oversight remains necessary, so 99% of the skills it looked at still need a person in the loop somewhere. The index publishes no breakdown for programming on its own, which means any programming-specific version of these numbers came from somewhere else.
The bigger change is where the first rung sits. PwC's 2026 Jobs Barometer compared US entry-level postings in 2025 against 2019 and found that among the skills newly appearing in the most AI-exposed quartile of occupations, 52% had been concentrated in experienced roles in 2019, against 7% in the least exposed quartile; the examples it names include strategic decision making, stakeholder management, process management and mentorship 6. That is an analysis of advert wording by a consultancy publishing its own index, not a measurement of what juniors are handed once hired. It still points somewhere specific. The first job used to reward producing code; it increasingly rewards specifying what the code should do and verifying that it does it.
Decide What to Do Inside the Degree
None of this answers whether to enroll for you, and an honest article does not pretend it can. What it supports is a decision about what to do once you are in the program, or while you are choosing it: spend deliberate time on the parts of the curriculum that build specifying and verifying, systems design, testing, reading someone else's code and finding what is wrong with it, not only the parts that reward writing a function from scratch fastest.
A short list worth checking against your own program:
- Does a required course make you review and correct code you did not write, not only produce your own
- Is there a project with a real stakeholder who can push back on a specification, not just a grading rubric
- Does anything in the program put you next to someone more experienced on a real problem, the way an internship, apprenticeship or rotation can
Which of those three your program already does is worth more to a decision than any national figure in this article, and the broader entry-level hiring picture is the wider context this narrower question sits inside.
Common questions
Is there solid data on unemployment specifically for CS graduates right now?
Not a figure this article could open and check. The source publishing recent-graduate unemployment broken out by major could not be reached, so no number from it appears here rather than being repeated secondhand. What is verifiable is a broader underemployment measure across all majors and a government youth-unemployment baseline. Treat any precise major-specific number you meet elsewhere with the same caution: ask which source and which release date it came from.
Does this mean programming jobs are safe from AI?
Not safe in the sense of unchanged, and not disappearing either. Indeed's index of almost 2,900 work skills judged 0.7% of them very likely to be fully replaced by generative AI and put the other 99% in categories that still require a person. Those are model judgments of what current tools can do rather than observations of hiring, and the index does not break programming out on its own.
Should I choose a different major instead?
This article does not have evidence that another major is a safer bet; the underemployment pattern it cites spans majors broadly rather than singling out computing. The more useful question is what inside any technical program builds specifying and verifying skills, not just production skills.
Is the tightening job market really about AI, or something else?
Both, and the honest answer separates them rather than picking one. Part of the current squeeze on new graduates predates generative AI entirely and reflects a broader hiring correction in graduate supply and entry-level demand. The measured contraction is also larger in the most AI-exposed occupations, though the economists who measured that call it a descriptive pattern rather than proof of cause.
What should I actually do differently in my CS program because of this?
Seek out coursework and projects that make you review, specify and verify code, not only write it, and look for any structured placement that puts you next to someone experienced. Those are the abilities AI-exposed entry-level postings have increasingly been written to ask for, on the evidence of how those postings changed between 2019 and 2025.
References
- 1. No Country for Young Grads: The Structural Forces That Are Reshaping Entry-Level Employment burningglassinstitute.org Supports the 52% underemployment figure across majors one year after graduation, and that the trend predates generative AI.
- 2. Labor Force Statistics from the Current Population Survey — series LNS14024887 '(Seas) Unemployment Rate - 16-24 yrs.' and LNS14000000 '(Seas) Unemployment Rate' data.bls.gov Supports the 8.5% against 4.1% July 2026 figures and the neutral baseline that youth unemployment runs roughly double the national rate, which is not evidence about AI.
- 3. Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence digitaleconomy.stanford.edu Supports the 22-to-25 employment divergence in AI-exposed occupations opening through hiring rather than separations, the authors' own descriptive-not-causal framing, and that the pattern is not confined to technology occupations or firms.
- 4. AI at Work Report 2025: How GenAI is Rewiring the DNA of Jobs hiringlab.indeed.com Supports that only a small fraction of assessed work skills are judged very likely to be fully replaced by GenAI.
- 5. AI at Work Report 2025: How GenAI is Rewiring the DNA of Jobs hiringlab.indeed.com Supports the four-way split of rated skills (40% minimal, 19% assisted, 40% hybrid, 1% full) and that hybrid means continued human oversight rather than full automation.
- 6. Two futures for jobs in an AI era — 2026 Global AI Jobs Barometer (global findings) pwc.com Supports the 52% against 7% split of newly appearing skills in 2025 US entry-level postings versus 2019, and the named examples of traditionally senior skills, which ground the advice about what to build inside the degree.
6 sources, numbered by first appearance. How Olive sources claims
General guidance for hiring teams. What works at one company and one volume may not transfer to yours.
Olive assesses how a person works with AI. It does not detect AI-written documents, and it never produces a score, a ranking, or a match percentage for a person. Candidates read the same report the employer reads.