No Old-School Coding, Pay for Your Own Tokens
CS students without Tokens should drop out immediately!
This “hot take” comes from Generative Software Engineering, a new course launched this fall by Jiang Yanyan, an associate professor at Nanjing University’s School of Computer Science.
Good grief—have computer science departments at Chinese universities become so competitive that students now have to “burn Tokens out of their own pockets”?
A related question on Zhihu quickly shot to the top of the trending list and has already attracted more than 400 answers.
But after taking a closer look, I found that most people in the comments actually support the idea.
And it seems to have struck a real chord:
Netizen A: That statement is far too conservative. I suggest universities cancel all classes and use the money saved on teaching hours to purchase Tokens and distribute them to students on a per-capita basis.
Netizen B: It sounds harsh, but I think it’s well said. At least it’s much better than the other hollow, useless, and flavorless takes.
Netizen C: Why stop at computer science?! I think students in every major should make “how to use AI” a compulsory subject, especially liberal arts students.
So, computer science students—and students in every other major—who don’t invest enough Tokens or know how to use large models proficiently…
Should they really all drop out?
Drop Out If You Don’t Pay for Your Own Tokens?
At the start of this academic year, Professor Jiang Yanyan delivered what was probably the most incendiary quote of the season:
I have a personal opinion: students who don’t have paid Tokens on hand should drop out immediately.
This remark came from his Generative Software Engineering class. Before interpreting it, though, we need to understand the context in which it was made.
To begin with, Professor Jiang had originally planned to secure some Token sponsorship. After all, asking students to pay out of pocket is, to put it mildly, somewhat difficult to justify.
But after seeing the actual number of students enrolled, he decided to let it go—there was no need to plead with vendors for Tokens, and doing so would be rather troublesome.
So he simply made “students pay for their own Tokens” one of the course policies.
What’s more, the course is about much more than simply having students pay for Tokens themselves.
Before making that statement, Professor Jiang first delivered a dose of reality, discussing the collective anxieties faced by CS students in this era.
The First Anxiety: AI Is Going to College for Me
The teacher uses Doubao to set the questions, and the students use Doubao to do their homework. Throughout the entire process, the only person who receives no training is the student.
Imagine that every course had open-book exams and all the questions were answered with AI, allowing you to score highly in every class. But by the time you graduate, you realize that although every course went well, you learned nothing—and AI replaces you first.
The Second Anxiety: I Studied Hard for Four Years, Only to Realize the University Lied to Me
There is another group of students who continue the attitude they had in middle school: they listen obediently to their teachers, pay attention in every class, and complete every assignment conscientiously.
But after graduating four years later, they discover that although they have genuinely learned some things, AI can do all of them too.
Then they become lost: What exactly did I spend these four years learning?
The bigger shock comes when they send out résumés during fall recruitment.
Professor Jiang says he has received countless résumés, including one that was “absolutely perfect”—with papers on the CCF journal list, accepted papers, and papers under review. It had everything a résumé should have.
But to him, it was a zero-point résumé: no GitHub link, no personal homepage, and not even a preprint.
Worse still, everyone’s résumé looked exactly the same: a mountain of internships, a mountain of awards, a mountain of papers, and a mountain of papers under review.
You and your competitors had become completely indistinguishable.
When even a “perfect résumé” is considered worth zero points, how exactly are you supposed to show a future advisor what makes you special?
It was against this backdrop of collective anxiety that Professor Jiang hoped to use the course to ease everyone’s concerns—or, more importantly, to help them rediscover a sense of joy: the joy of vibe coding.
He says he has found a great deal of joy in vibe coding. In the process, students can understand how software engineering works, build interesting things, and make their résumés look better.
For him, having a published paper might even count against a student.
The rules of the course are also refreshingly unconventional. Professor Jiang listed three policies:
- No old-school coding;
- Students pay for their own Tokens;
- But Tokenmaxxing is not required.
“No old-school coding” is easy enough to understand: stop writing code the way people did ten years ago. The times have changed.
Professor Jiang jokes that everyone in the room is an expert at old-school coding.
After all, students still have to take exams, complete programming tests, and compete for recommendation-based admission to graduate school. If they are really locked in a room with all AI tools disabled, the things that can still save them are LeetCode and muscle memory (doge).
But he believes this will eventually change.
Generative Software Engineering is his attempt to open up a small breach in the current system.
The question this course aims to address is: As AI gets faster and faster at writing code, what work is left for humans to do?
In this course, students are expected not only to “know how to code,” but also to learn how to break tasks down for AI, choose models, control costs, and verify results. No matter how quickly a model generates code, it is useless if people don’t know whether the code is correct.
“Tokenmaxxing is not required” means you shouldn’t treat Tokens like water and burn through them recklessly. You need to use them efficiently.
Although Professor Jiang requires students to pay for their own Tokens, he also recommends a model with a sufficiently high cost-performance ratio:
DeepSeek-V4-Flash, of course!!!
In this course, you might even hear the instructor suggest doing your homework during “Liang Wengu” time…
Aside from using a smarter model to generate the slides, he uses DeepSeek-V4-Flash for every demonstration in class and for every asset generated.
He also recommends that students top up their DeepSeek accounts with 100 yuan. Students who genuinely have difficulty affording Tokens can contact him, and he will help them find a solution.
What else is there to say? After hearing all this, I actually think this teacher sounds pretty sincere…
Besides, paying for your own Tokens seems to have become an unavoidable reality for computer science students.
One netizen commented:
Although it is reasonable in a certain sense, I suggest that any computer science department that doesn’t buy Tokens for its students should be shut down immediately.
That actually makes a lot of sense~ If Tokens are the “laboratory consumables” of the AI era, then chemistry departments provide the reagents, and physics departments purchase the equipment with university funds.
Why, then, do computer science students have to pay out of pocket?
And not every student can afford to burn through Tokens.
These days, major Silicon Valley companies set Token allowances and caps on AI-related spending for their employees. As for AI computing accounts at Chinese universities, some institutions that moved quickly have already put them in place.
Jiang Yanyan: A Teacher Who Loves Memes and Telling Students to “Drop Out”!
By this point, everyone must be wondering:
Who exactly is this teacher who casually tells students to “drop out”?
Jiang Yanyan is an associate professor and doctoral advisor at Nanjing University’s School of Computer Science. He graduated from Nanjing University with a bachelor’s degree in 2011 and a PhD in 2017, after which he stayed on as a faculty member.
He also has an exceptionally impressive academic background:
As a student, he won gold medals at the ICPC regional contest twice and made it to the 2009 ICPC World Finals.
In 2016, he was named one of Nanjing University’s Students of the Year. In 2018, he won the CCF Outstanding Doctoral Dissertation Award for his work Research on Shared-Memory Access Dependencies in Concurrent Programs.
After becoming a professor, Jiang Yanyan’s main research areas have been systems software and software automation.
According to Nanjing University’s School of Computer Science, he has received CCF A-category conference paper awards in operating systems and software engineering five times, including the sole Best Paper Award at ICSE 2021 and the Best Paper Award at SOSP 2023.
His research focuses on systems software and software automation, and he has recently shifted toward Agentic AI. He is currently also the Tech Lead at Cosmic Dawn AI.
However, in the course description, this prominent scholar describes himself as “an associate professor, but a temporary worker.”
(Institute of Computer Software)
That’s quite a sense of humor.
Many people who know him, moreover, discovered him not through his papers, but through Bilibili.
Jiang Yanyan has made courses such as Operating Systems: Design and Implementation, Fundamentals of Computer Systems, and An Introduction to Software Engineering Research publicly available for years. His videos have accumulated more than five million views in total.
He also maintains a wiki for each course with extraordinary care, open-sourcing all the lecture notes, labs, and slides.
His students call him a “legendary teacher,” and he has even received Nanjing University’s “Students’ Favorite Teacher” award.
In Jiang Yanyan’s classes, he often explains concepts while opening a terminal to write code live, casually throwing in a few jokes that only programmers can understand instantly.
His course homepage likewise feels like one giant meme in the making.
For example, the first lesson of the semester begins with a warning:
Without sufficient effort, you will fail programming.
He then tells students to put the most important thing first:
All course information is posted on the course homepage. RTFM.
As for whether students need to attend class, he is fairly laissez-faire:
Attendance is not required. Spend your time on more effective pursuits.
This style has been taken to an entirely new level in this year’s Generative Software Engineering course.
“No old-school coding” and “students pay for their own Tokens”? Those barely even count as extreme by Professor Jiang’s standards.
He has also described the class as a “performance,” hoping to use AI to continuously modify, slice up, and maintain the course, ultimately achieving the course’s “digital immortality.”
The entertainment value is truly off the charts.
Even his Bilibili account name is:
The Green Advisor Has Forgiven You.
The meme comes from an article he wrote titled How the Green Advisor Got His Hat: The Origins and Development of the Academic Great Leap Forward, which primarily discusses the academic evaluation system in higher education.
His personal homepage is also quite interesting: it is designed to look like a Linux terminal.
The page uses ASCII characters to spell out his name. Visitors can also enter commands such as bio, help, and tree just as they would in a command line to view his background, papers, and team information.
Even his office number is written in binary.
Of course, although Professor Jiang can be quite “abstract,” he is also a man who loves writing “dropout letters.”
He previously wrote A Dropout Letter to the Chosen Ones.
He also wrote The Things About Pursuing a PhD, an account dedicated to his own five years as a doctoral student.
In it, he tells students directly:
If you can’t even stand up to your advisor face-to-face, you should think very carefully about whether pursuing a PhD is right for you.
In that light, perhaps it really isn’t surprising that this new Generative Software Engineering course starts off by telling students to drop out…
Is it?
References:
[1] https://www.zhihu.com/question/2075868388124004913
[2] https://www.bilibili.com/video/BV1pb8o6yE8f/?spm_id_from=333.1387.homepage.video_card.click&vd_source=954d569b809ce6c1899872814368cb8a
[3] https://jiangyy.github.io/
[4] https://jyywiki.cn/