At some point last year, almost every engineering, software development, or data science student opened Copilot, ChatGPT, or Claude for the first time. No professor asked them to. No course syllabus required it. They did it naturally, because they had a problem and the tool was right there.
The numbers say that 85% of 2026 graduates used AI during their studies, and only 28% received formal instruction on how to do it. That means the vast majority figured it out on their own, without guidance or framework. Not exactly the most encouraging picture for higher education, is it?
The Problem Isn’t That Students Are Using AI
Students got there on their own. AI usage among college students jumped from 66% in 2024 to 92% in 2025. In tech-focused programs the adoption is even more pronounced: anyone studying software development, data science, or systems engineering has daily access to tools that generate code, documentation, and technical solutions.
What was missing was structure. Nearly half of 2025 graduates felt unprepared to apply for entry-level roles, and only 51% believed they had sufficient AI skills for the jobs they applied to. Employers are increasingly treating AI fluency as a baseline requirement. Universities and technical programs are racing to close that gap, with very different speeds and resources.
Institutions didn’t fail by banning AI. They failed by ignoring it until it was unavoidable.
What Curricula Still Haven’t Solved
Curriculum changes at universities move through department committees, college curriculum councils, university senates, and accreditation review. The process exists for legitimate reasons. But the generative AI landscape doesn’t operate on 18-month cycles. The tools that formed the basis of AI instruction in 2023 have been superseded multiple times. The models students learn to use in a course approved in January 2025 may function in substantially different ways by the time that course runs in fall 2026.
That’s the core tension. It’s not bad intent. It’s that the speed of change and the speed of academia are structurally incompatible.
Research shows that students who use AI heavily without guidance develop weaker skills, produce less original work, and in some cases experience heightened anxiety tied to dependency on automated tools. AI supports learning when students engage with it critically. It undermines learning when they use it as a substitute for thinking. And the problem starts before the classroom: more than two-thirds of urban teachers haven’t received any AI training. The classroom can’t give what the instructor doesn’t have.
Universities must look beyond their campuses. Because commercial LLMs are often structurally opaque, traditional curricula face practical challenges in teaching students to critically evaluate AI outputs without direct collaboration with the technology sector.
Who’s Already Found Another Way
Some institutions and programs are redefining the model from within, without waiting for institutional consensus.
Stanford University launched “The Modern Software Developer,” a course where students generate everything through AI without being allowed to write code manually — a radical departure from traditional pedagogy. Anthropic, in partnership with CodePath, the largest provider of collegiate CS education in the US, redesigned coding curricula around AI tools. In a fall 2025 pilot, over 100 students used Claude Code to contribute to real open-source projects. CodePath’s CEO put it directly: “We now have the technology to teach in two years what used to take four.”
Bachelor’s AI programs in the US grew 114% from 2024 to 2025, from 90 to 193 programs. Northwestern University’s new AI major, launching in 2026, is among the latest examples of institutions integrating technical training with ethical scrutiny. The movement isn’t just technical. It’s structural.
What Should Change — and What Shouldn’t
There’s an easy conclusion to draw here: if AI generates the code, teaching programming no longer makes sense. That conclusion is wrong, and it’s worth saying clearly.
Many introductory AI courses emphasize algorithmic techniques and code-heavy assignments, while critical analysis of AI is reserved for the end of the semester, if time allows. University education should teach fundamental concepts, discuss whether and when an AI solution is appropriate, and explore the ethical implications of AI use.
What changes isn’t the need to understand code. What changes is the order in which it’s learned and the hierarchy of skills. You don’t start with syntax anymore. You start with judgment: understanding what makes a good system, what makes good code, what makes a sound architecture decision. AI can write the function. It can’t tell you whether it was the right function to write.
Nearly all employers across all business sectors now say they are or soon will be expecting employees to possess AI competencies. They are searching for programs that produce graduates with AI skills — and they are willing to pay a premium for those with such skills. The market has already moved. The question is whether the curriculum follows.
Our Take: Tech Education Has a Speed Problem, Not a Direction Problem
Academia isn’t heading in the wrong direction. It’s moving too slowly for a change that already happened.
CS and software development students who enter the job market in 2026 with the judgment to evaluate AI-generated code, the ability to design systems, and an understanding of the limits of the tools they use have a real advantage. Those who graduated with four years of syntax and no structured exposure to AI are at a disadvantage — not because they’re worse developers, but because they trained for a job that no longer exists in the form they knew it.
The question institutions should be asking isn’t “do we allow AI?” It’s “what market are we preparing our students for?” If the honest answer is “the one from five years ago,” the change is overdue.
Sources: Stanford University / Medium, High Ed Insights / Substack, Scholaro: AI Impact on Higher Education 2026, DemandSage: AI in Education Statistics 2026, Science.org: Higher Education Must Bridge the AI Gap, AAC&U: Institute on AI, Pedagogy, and the Curriculum
