The First Generation of University Students Studying AI Are No Longer Smiling
The year 2026 is shaping up to be dominated by AI. Programmers are being laid off, one-person companies are booming, and lobster-like intrusions are occurring globally. The hype surrounding AI is causing increasing anxiety among ordinary people. This article explores the realities faced by AI students, revealing a gap between external expectations and their actual experiences.
Is AI truly the future? If so, what about my job? Many regret not choosing an AI-related major in college, feeling they’ve missed the biggest opportunity of our time.
To outsiders, those studying these majors seem to have it made – a perfect “success story” script, constantly researching how to build large models, mastering AI better than anyone. We, on the other hand, feel like ignorant foreigners, just tagging along and benefiting from AI.
So, what’s life actually like for those who have secured a ticket to this new era?
Driven by curiosity, I spoke with several undergraduate, master’s, and doctoral students studying AI at different universities. Surprisingly, within these ivory towers, so highly praised by the outside world, there wasn’t an abundance of confident pioneers. They’re actually just like everyone else, stumbling and groping their way forward within the constraints of reality.
First, the school’s curriculum hasn’t kept pace, and they’re never taught *how* to use AI in class.
This is understandable, though. The school expects them to *create* AI, so naturally, they must first teach the foundational knowledge. There’s nothing to criticize there.
However, the reality is that many university AI programs were created by splitting off from information or computer science departments, and the curriculum is still in transition. Some lean towards electronic communications, while others still rely on the foundations of old computer science programs.
Li Mo, a student from a 985 university in Anhui, complained that their program was the first year AI separated from the School of Information, and the curriculum hadn’t been fully revised. They’re required to take “Signals and Systems” from the communications program, but aren’t required to take “Compiler Design” and “Operating Systems,” essential computer science courses.
Most agonizingly, they also have to take “Electromagnetism,” “Quantum Physics,” and a host of other physics courses… which are difficult and useless for computer science.
Besides theoretical courses, programming isn’t taught perfectly either.
Those familiar with AI know that Python is currently the most widely used programming language in the field. However, a student who calls himself “Late 9 AI” said that their curriculum still focuses on C++, which creates a disconnect between “learning” and “doing.” They routinely use Python for projects, but are forced to write C++ for coursework and exams, a holdover from the neighboring computer science department’s curriculum.
Learning and applying different things is a test from the AI gods, perhaps.
Another issue is that AI programming has become so powerful that it’s making waves. News of layoffs is constantly circulating on the internet, and Dario, the founder of Claude, has revealed that the vast majority of code is no longer written by employees.
But schools certainly aren’t keeping up with the times. In the survey, I learned that there aren’t yet any courses specifically on AI programming; everyone is sticking to traditional programming methods.
Teachers, while explaining PowerPoint slides, might occasionally mention, “GPT and Claude are very popular right now, many people use them to write code, so what does this tell us? What students really need in the future is creativity.” And then they’re done. If you want to truly learn AI programming techniques, you have to follow the tech world and watch videos on your own.
More than one student mentioned that some of the knowledge and technology stacks taught at school are already outdated. The teacher’s role is to open the door to this field, but further exploration is up to the students themselves.
I think that’s reasonable. I graduated many years ago, and the school’s update cycle is still so slow, so I’m relieved.
Until Pony, from a 211 university in Zhejiang, told me, “The teachers don’t write code, we do it all.” I suddenly understood – students are the best agents.
Teachers’ attitudes towards AI are also quite divided.
Zhang Fa Cai, a student from a 985 university in Hunan, told me he’s studying network security. In terms of using AI, teachers in the network security field explicitly discourage it, believing that many operations must be completed by hand to truly internalize the principles, especially in the security field, where handing operations over to AI is a vulnerability in itself.
However, teachers in the web development field are the opposite, actively suggesting the correct use of AI in class, saying it can save a lot of work in web development.
To be fair, both sides of the teachers are right. Network security requires practitioners to have a near-muscle-memory grasp of underlying principles, a step that cannot be skipped. And web development involves a large amount of patterned work, which AI can indeed significantly improve efficiency.
The problem is, no one can guide students to form their own judgment system: when to use AI, when to rely on their own hard work, and how to use it without ruining their fundamentals. It’s all up to the students to figure out on their own.
This could lead to a problem: those who can use AI use it as a tool and learn faster, while those who can’t throw everything at AI and end up with nothing to show for it after a semester.
As a result, the students no longer pretend, you don’t teach me, so I’ll learn myself. When asked about the classroom situation, Xiao Xue, a student from a 985 university in Shanghai, said, “Everyone rarely listens to the teacher now, they’re all buried in learning. What the industry needs is different from what the school teaches.”
And because the use of AI is becoming more and more convenient, the AI usage rate in coursework is already very high. “Now, writing course projects basically involves AI participation throughout the entire process, there’s no one who doesn’t use it.” Zhou Kai, a doctoral student from Shanghai, told us. The same answer was received from students in other grades.
Interestingly, everyone mentions how “convenient” it is to use AI, but no one feels they’ve fully mastered it.
Zhang Fa Cai said that over-reliance on AI will lead to a loss of control over the project, and the code will become a mess. The time saved writing code is all spent reviewing it.
“It’s just that what’s written isn’t what you want, and fixing it takes a lot of time. You still have to have development experience and type code yourself first before you can truly master AI.”
Doctoral student Zhou Kai said that he spent 200 million tokens during his winter break, and his final feeling was: it’s okay to let AI write simple scripts directly, but you still need to design the high-level architecture yourself. “Because after chatting with AI for a long time, you’ll find yourself getting dumber. The summaries AI gives you look good, but you can’t see the mistakes, and it’s actually slower than doing it yourself. Many experts dare to use AI confidently because they already know what’s right and what’s wrong.”
Xiao Xue, who is preparing for job applications and brushing up on questions, directly began to complain: after using AI for a long time, she can’t write code herself, and she can’t solve any problems.
Everyone is frantically using AI, but no one feels they’ve found the definitive answer.
And reality is a different beast altogether.
Today, the requirements of internet companies for new graduates are bizarre. Whether or not to “hand-code” (writing code questions without any tools) during interviews has become a matter of speculation. Some still insist on traditional programming methods, while others have evolved to the AI programming stage.
Teachers always tell you to use AI, but you might not be allowed to use it during the interview, which is like being asked to derive the Pythagorean theorem on the spot during the college entrance exam. The tears only you know.
But those we’ve talked to, whether they’ve figured things out or stumbled, at least have the energy to think.
For more related majors, the real state of affairs is, who cares?
Those preparing for graduate school are studying, those securing research positions are busy with competitions and research, and those looking for jobs are scrambling to prepare. All this talk about new AI tools and the latest AI developments just feels noisy to them.
Zhang Fa Cai bluntly said that reality is too heavy. “After finishing my GPA, I realized in my junior year that research experience doesn’t just appear.”
Take OpenClaw, the most representative product of 2026, for example. It exploded online, and the media covered it extensively, and major companies followed suit without hesitation.
But AI-related students are mostly indifferent. “I feel like I can’t earn back the money for the tokens.” Zhang Qiu said.
And Pony told us, “There’s basically no splash among my peers, maybe just some teachers who want to play with it.”
Zhou Kai is more concerned about security, it’s too dangerous, afraid of being injected with attacks.
A product seen by the outside world as “changing the world” isn’t so popular with those who actually study AI. I’ve summarized two voices: one sees the essence, “Lobster is essentially a roadside one,” and the other is pressure, everyone has more urgent matters.
Pony also told us that people in his lab react half a beat slower than the outside world.
“Most of my peers are still traditional. When everyone was using LLMs, we liked to code by hand. Now everyone is using AI coding, and we still love LLMs. We’re too lazy to try new tools, and we’re more concerned about whether we can publish papers. My workflow uses vim, hand-coding in the terminal, and VSCode’s Copilot—these old things are enough, why should I switch to new ones?”
From undergraduate to doctoral studies, the outside world is enthusiastically discussing how AI is changing the world, while those inside are thinking about how to get through what’s in front of them. Because the acre of land in front of them is their world.
Therefore, we can summarize the current situation in the AI circle: the old timers are busy directing others to embrace AI, the young ones are busy with a promising future and have no time to play with AI, and the middle ones are playing with AI while shouting “the future is here.”
The rest are ordinary people who are forced to accept AI’s influence.
Finally, when asked if they would continue in this field and what direction they would take, most people said they were clueless.
However, Zhou Kai still offered his perspective. “AI is leveling the playing field for everyone’s engineering skills—if Code Agent allows everyone to write code, what’s your advantage over others?” He has recently been consciously learning communication skills, because he believes that in this era, clearly describing something is the most valuable skill.
But more people express confusion, confused about their current coursework, interviews, and eight-legged essays, confused about the rapidly sweeping AI wave.
Everyone is in the same water, choking and learning to swim.