Editor: Aeneas, so sleepy
[New Intelligence Prime] U.S. coders are experiencing a 'massacre,' with the employment rate plummeting by 27.5%, and nearly one-third of jobs disappearing. CS graduates of 2026 have no path left. A programmer of many years said: This profession is disappearing, may we exit with honor and have a blast.
The species of U.S. coders is gradually going extinct.
This is not some alarmist prophecy, but an unfolding reality.
Global layoffs caused by AI reached 1.17 million in 2025, the highest record since 2020.
Computer science graduates of 2026 face a scorching water-and-fire situation right after graduation—fundamentally unable to find a job!
And data from the U.S. Bureau of Labor Statistics shows that the employment rate of U.S. programmers has already plummeted by 27.5%.
In other words, nearly a third has been cut.
What to do? "This brutal pleasure will eventually end with brutality."
U.S. coders are nearly extinct?
Today in the United States, the employment rate for programmers has plummeted.
The labor bureau's data shows a drop of 27.5%.
And Stanford University research found that since the end of 2022, with the popularization of AI tools, the employment rate for programmers aged 22 to 25 has decreased by nearly 20%. (Address of the paper "Canaries in the Coal Mine?": https://digitaleconomy.stanford.edu/publications/canaries-in-the-coal-mine/)
Researchers analyzed wage records from ADP, the largest payroll company in the U.S., tracking millions of employees working at tens of thousands of companies between 2021 and July 2025.
Data shows that the employment situation for young and older developers was consistent until the end of 2022. But since then, the two have diverged—young developers have started to lose jobs, while older developers have not.
According to statistics from a U.S. consulting firm, the impact of AI-induced layoffs on the U.S. this year is second only to the pandemic.
The report released by this company shows that AI has directly or indirectly led to nearly 55,000 job losses in the U.S. this year!
A programmer wrote in his blog: "Why is everyone I know being laid off?"
In the U.S. tech industry today, it's really hard to work safely. The shadow of layoffs and AI is looming over every programmer.
How to name this disaster. The dot-com bubble burst? The Great Recession? The Unicorn Massacre? Or CrashGPT.
This programmer wrote: Meta laid off thousands, Google froze hiring, this is a slow collapse of a collective illusion, the FAANG dream is rotting from within.
AI Layoff Cataclysm
AI is no longer a tool to enhance productivity, but a role that directly replaces programmer positions.
Stanford, Toronto CS Graduates
Job Hunting in a Dead End
Moreover, AI has left young CS graduates no way out.
Stanford CS graduates, upon graduation, found the situation completely different from three years ago, which made them very angry.
Because they couldn't find jobs, many had to choose to study for an extra year of graduate school. But three years ago, many had already found jobs before graduation.
Azka Azmi graduated from the University of Toronto's computer science department this spring and still hasn't found a job.
The more she looks for a job, the more frustrated she feels, because in this process, she has almost no chance to talk to a real person!
AI is everywhere, all companies are using AI to replace real people for recruitment, all you can do is adapt to this world where machines talk to each other.
Once, CS was the most sought-after major in everyone's eyes, with salaries often reaching a million, plus high benefits and a fun work environment.
But now, due to AI, economic uncertainty, and a large number of CS graduates entering the workplace, these legendary perfect positions have disappeared with a snap.
Azka Azmi said many students rely on internships or co-op opportunities to find jobs, but now only about one percent of applicants get a response.
Getting a Master's, Only to Find It's Harder to Find a Job
In the spring of 2024, Elliot Chen obtained his CS degree from the University of Toronto and then submitted hundreds of resumes.
He frustratingly found that there were very few opportunities for fresh graduates. Many positions required at least one year of non-internship work experience, which most fresh graduates do not have.
Many people can't even pass the resume screening stage.
Because of the unsuccessful job search, Elliot Chen decided to continue pursuing a CS master's degree to stand out.
Result: halfway through his master's, he found that the employer responses he received were even fewer than when he was an undergraduate!
A CS doctoral student found that undergraduates in this major are extremely panicked about the current job market, even showing mental health issues.
"The competition is very fierce, many environments have become very unfriendly. These kids do everything. They go beyond the limits of anyone before. It's cruel for everyone."
Chrisee Zhu also felt that her classmates were unusually anxious.
In group courses, they often daydream, unable to contribute, and instead focus on job applications and programming practice, preparing for technical interviews.
Karpathy
Programmers Are Experiencing a Magnitude 9 Earthquake
Feeling AI fear is not just for novices, but also for big shots.
Just now, former Tesla AI director and OpenAI co-founder Karpathy said he was shocked by powerful alien technology!
The "powerful alien weapon handed to humanity" he spoke of is AI.
And he said bluntly—
I never felt so behind as a programmer.
He deeply feels that as programmers contribute more sparse and fragmented code, this profession is undergoing drastic changes.
He regretfully stated that if he could properly connect the new tools that have emerged over the past year or so, his abilities could have been enhanced tenfold; but if he cannot master this enhancement, it is a skill defect.
Karpathy concluded: Now, we face a completely new programming abstraction layer that we must master (it is superimposed on the original tech stack), involving agents, sub-agents, prompts, context, memory, patterns, permissions, tools, plugins, skills, hooks, model context protocol, language server protocol, slash commands, workflows, IDE integration...
More urgently, we need to build a global mental model to understand the pros and cons and pitfalls of these "intelligent entities" that are inherently random, error-prone, elusive, and constantly evolving.
And now, they are suddenly intertwined with traditional rigorous software engineering.
Karpathy finally exclaimed:
Obviously, a powerful "alien tool" has been handed to us, but there is no instruction manual!
Everyone can only figure out by themselves how to hold and operate this tool.
In his words, this magnitude 9 earthquake is shaking the entire industry! Roll up your sleeves, don't get left behind.
Anthropic engineer Boris Cherny said in his comment: He now feels this way every week.
Whenever he handles a problem manually, he finally finds: Claude might be able to handle this.
Karpathy said he has similar experiences.
When you wave this weapon around everywhere, it might go off; but if you can hold it just right, a powerful laser will shoot out, directly melting your problem!
In the Garbage Code Made by AI
Programmers Fight for Survival
"AI won't take your job, but people who use AI will."
Since October 2023, this sentence has almost become NVIDIA CEO Jensen Huang's signature mantra.
In the tech wave of the past two years, this maxim has been quoted countless times, becoming the Sword of Damocles hanging over every developer.
Time has come to the end of 2025, and the prophecy seems to be coming true in a confusing and contradictory way.
On one hand, a survey of the entire industry by Google shows that as many as 90% of tech positions are now using AI tools. In 2024, this proportion was only 14%.
On the other hand, giants like IBM and Amazon are laying off employees wildly, while the survivors are being drowned in a "tsunami of technical debt" full of bugs and vulnerabilities created by AI.
Now, what we have to face may not just be a reshuffle of the job market, but a crisis about the essence of software engineering itself.
The Massacre in Progress: The Extinction of "Code Farmers" and the Survival of "Developers"
As mentioned before, if you are a fresh graduate of 2026, you may be facing the most severe job market in decades.
According to the "2026 Employment Outlook" released by the National Association of Colleges and Employers (NACE), employers' pessimism has reached its highest point since 2020.
https://www.naceweb.org/research/reports/job-outlook/2026/#data
Here is a highly ironic data contrast that reveals the precise strike of AI on the industry.
According to data from the U.S. Bureau of Labor Statistics, between 2023 and 2025:
- "Programmers": Employment rate plummeted by 27.5%. These jobs typically involve writing code according to established specifications, with independent and highly structured work.
- "Software Developers": Employment rate only slightly decreased by 0.3%. These jobs are more focused on design, architecture, and solving complex problems.
At the same time, positions for information security analysts and AI engineers have seen double-digit explosive growth.
Jamie Grant, senior associate director of Career Services at the University of Pennsylvania, has a sharp analysis. She warns students pursuing software engineering positions:
Now, positions are no longer just about writing code.
Employers require higher-order thinking skills, control over the software development lifecycle, and skills that AI cannot replace—such as understanding customers' vague requirements.
AI Code Collapse: The Mythical "Shit Mountain" Generator
Since "programmer" positions are decreasing, is it because AI is doing the job better?
No. The truth is startling.
AI software company CodeRabbit recently released a shocking report, pouring a basin of cold water on the industry blindly worshiping AI programming:
Code written by AI is simply a mess full of bugs.
CodeRabbit analyzed 470 code merge requests (Pull Requests) and reached a quantitative conclusion:
- Human code: On average, each request contains 6.45 issues.
- AI code: On average, each request contains 10.83 issues.
In other words, the error rate of code generated by AI is 1.7 times that of humans.
https://www.coderabbit.ai/blog/state-of-ai-vs-human-code-generation-report
More worrying is the nature of the errors. The proportion of "serious" and "major" issues in AI-generated code is extremely high.
Although AI is twice as good as humans in spelling and grammar, once it makes a mistake, it rises to that kind—deep logical fallacies, lack of functional correctness, and code readability disasters.
CodeRabbit's report points out that these problems are accumulating like a snowball into a huge "long-term technical debt."
In addition, research from security company Apiiro added another blow:
Developers using AI produce security problems that are ten times those of their peers who do not use AI.
Because AI often "loses intelligence" when handling passwords and sensitive information, leading to the leakage of protected information.
Bain & Company stated bluntly in its September report:
Although programming is one of the first areas to deploy generative AI, "cost savings are not significant," and "results have not met the hype's expectations."
The Absurd Reality of Cleaning Up AI's Mess
This "high output, low quality" characteristic is fundamentally changing the daily work of engineers.
David Loker, AI Director at CodeRabbit, said: AI does accelerate output, but it also introduces predictable, measurable weaknesses.
This change forces human developers to take on a new role—cleaning up after AI.
In July, a study from METR revealed a counter-intuitive phenomenon:
For experienced developers, AI tools actually slow down their progress.
Why? Because programmers are forced to become full-time "nitpickers."
They need to examine the seemingly perfect but actually flawed code generated by AI with a microscope. As long as one hidden logic bug is missed, the entire system could crash.
But this doesn't mean we should abandon AI.
Jamie Grant compares AI to an "exoskeleton":
Imagine, it allows you to easily lift 1000 pounds. It should be an enhancer for your work, strengthening your higher-order critical thinking.
NACE data supports this view:
61% of employers say they are not simply using AI to replace entry-level positions, and 41% plan to use AI to enhance these positions.
Broken Promotion Ladder: Workplace Newcomers in the Bottom Hell
At the same time, this transformation brings a more profound crisis: How should the new generation of engineers grow?
In the past, junior engineers honed their skills by doing simple, task-oriented "grunt work" and gradually grew into experts who could stand on their own.
But now, these jobs are taken over by AI.
https://www.signalfire.com/blog/signalfire-state-of-talent-report-2025
Now, fresh graduates are forced into a dead cycle of "which came first, the chicken or the egg":
If basic work is all completed by AI, newcomers need to be competent for higher-level work from their first day on the job.
But without the exercise of basic work, how can they acquire the ability for advanced work?
Swipe left and right to view
In response, Mike Roberts, founder of Creating Coding Careers, warns that many companies are short-sighted, only focusing on next quarter's performance and unwilling to invest in training newcomers.
If you don't train new entrants in the market, you will eventually be unable to hire mid-level backbone, which is very short-sighted.
AI Is Not on the Negotiation Table
By 2026, whether for battle-hardened architects or fresh graduates, the rules have changed.
The era of surviving by "memorizing algorithms" or "piling up code volume" has completely ended.
As Jamie Grant said, students and newcomers must recognize where AI cannot help:
At critical moments on the negotiation table or expanding client relationships, AI may not be by your side. You still need to demonstrate your highest level of personal capability.
Future engineers are destined not to be just the "code farmers" silently tapping keyboards in the corner.
You must evolve, you must become a strategist who understands the business, a rigorous security reviewer, and the super driver who can tame the "bug generator."
Technology has not eliminated humans, it just cruelly deprived the mediocre of the right to survive.