Today I came across an article from 36Kr, and I recommend everyone take a look. Although it contains elements of anxiety-mongering and exaggeration, we have to admit that this is the current reality and will be for a long time to come. 2026 is just the first step.

Even if you embrace AI, you may still be eliminated by AI, and this process is accelerating. In the face of AI, all white-collar positions have no moat.
The original text is as follows:
The first batch of big tech workers cut by AI: high salary, high performance, high ranking
"Now the company has a (layoff) list, and you're on it." One day in mid-May, Lin Yue was called into a meeting by his team lead, who got straight to the point.
Lin Yue's first reaction was calm; he had expected it. As early as March or April this year, rumors of layoffs had been circulating inside some internet companies. Since the start of the year, China's major internet companies have been aggressively pushing AI efficiency through token competitions, training sessions, and hidden assessments—everywhere. When everyone is swept up in an "all in AI" movement, the consensus is clear: "Layoffs will definitely happen."
But standing at the HR door, he still had a moment of emotional breakdown: his hands started shaking, he hesitated for a long time, thinking about how to start, how to adjust his demeanor and expression. "I never want to go through this again."
Lin Yue had a monthly salary of 25,000 RMB. A year ago, fresh out of undergraduate, he joined Ctrip as a backend engineer—at the time, he was extremely lucky. The bonus era of internet recruitment was over; out of thousands of resumes, Ctrip admitted fewer than 500 people. But he entered the company's most profitable hotel department, responsible for writing code for commercial products.
But now, a junior programmer with a salary of 25,000 RMB and only one year of experience—who else would they cut? First, the severance cost is low; second, compared to veteran employees who are more familiar with the overall business, new hires tend to use AI less efficiently. "With business experience as a foundation, veteran employees know better what to do with AI and what impact it will have," Lin Yue said.
In a paper titled "Canaries in the Coal Mine?" Stanford University uses "canaries" to describe young people just entering the workforce. Its research shows that since ChatGPT became widespread in 2022, employment among the youngest workers has dropped significantly. By September 2025, employment among software developers aged 22-25 had fallen nearly 20% from its peak at the end of 2022.
In the past year, AI has made everything more intense. Ctrip was once a famous "internet retirement home": programmers started work at 10:30 AM, had a two-hour lunch break, and could leave at 7 PM on the dot, with the main app iterating every two weeks. But shortly after Lin Yue joined, AI coding capabilities exploded, and the pace became so intense that the app iterated once a week, "working until 10:30 every night."
But this acceleration wasn't due to explosive business growth; "it's because if you don't find things to do, your department becomes marginal, and marginal departments get cut," Lin Yue told 36Kr. In the end, he still couldn't avoid being "cut."
However, the "culling" might be indiscriminate.
Cang Shu never expected to be among the first names on the layoff list.
One Friday in May, half an hour before work, "the department suddenly called an all-hands meeting, and HR directly announced the results, telling everyone about it."
Before joining Meituan, Cang Shu was an SSP (Super Special Offer) campus hire at ByteDance, entering with a high salary and eventually becoming the highest-paid employee at the same level within his team. After jumping to Meituan, almost all core projects within the team were placed in his hands. This year was supposed to be Cang Shu's promotion node.
In this wave of layoffs, the protective barriers of "top performers" and "high rankings" have all failed. In the team next to Cang Shu's, two employees who had received "above expectations" performance ratings last year were let go. In the end, Cang Shu's own team was almost entirely "wiped out"; "the team still exists in name, but practically no one is left."
When Lin Yue learned he was laid off, he discovered that two frontend engineers he often worked with "had their profile pictures grayed out at some point without him noticing"; a large user growth group at Meituan, originally with hundreds of members, now only had about half left; Alibaba's Amap and Fliggy businesses were also in turmoil.
"630" became a buzzword on social media. It marked the end of the first quarter when AI truly entered internet workplaces on a large scale in China. From late June to mid-July, both the usual timing for personnel changes at many companies and the commonly set "last day" for this wave of layoffs.
The bellwether Silicon Valley has already started layoffs, characterized by large batches and scale. In May, Meta announced layoffs of 8,000 people, with 7,000 transferred to AI departments, making it the most volatile among Silicon Valley tech companies. Executives admitted that "company morale is at its lowest in 20 years." Earlier, Amazon announced layoffs of 16k white-collar positions, redirecting the savings to AI.
Before the previous round of layoffs in 2021, major Chinese internet companies were aggressively expanding their boundaries, rapidly establishing one new business after another, quickly recruiting a batch of people and then quickly eliminating them.
But the internal logic of this year's layoffs is not so singular. AI efficiency improvements, sluggish growth or deep competitive struggles of old, large businesses, and the cash pressure from investing in new AI businesses are all intertwined. Many people who were told to leave find it hard to weigh these factors.
The author of "Hassabis: The Brain of Google AI" said that like Oppenheimer, who created the atomic bomb but couldn't control its use, scientists pursuing truth are also "destroyers of all things": our work, our way of thinking, even our survival could be "destroyed." Ten years ago in Seoul, South Korea, AlphaGo brought that initial destruction to human Go player Lee Sedol. Ten years later, from Silicon Valley to Beijing, that destruction is spreading again.
For large companies, AI is a ticket to new businesses like large models or AI applications. But whether new businesses will succeed and when, no one can say for sure. Facing old businesses that are no longer growing, large companies have no choice but to be more resolute in improving efficiency in every certain and uncertain direction, and then laying off.
When Lin Yue confided in a friend about his layoff, he was comforted: "It's okay, we'll all have our day. Yours just came a little earlier." But perhaps more important than self-consolation is: after being replaced by AI and laid off by big tech, how should people choose and act?

Anxious top management, escalating middle management, desperate frontline
"What used to take two months to make a product demo at ByteDance, we can now do in two weeks," a former ByteDance product manager, now an executive at an AI startup, told 36Kr. With tools like Claude Code and Codex, his team can now make a demo in three hours and validate an idea within a week.
"A product (manager) is like a CEO," he said. The organizational structure can be drastically compressed, with far less information transmission loss than in big tech—perfect "entropy reduction."
While startups move quickly with AI, do internet giants look back at themselves and feel like slow-moving behemoths?
Statements from the highest levels of big tech often serve as signals.
In March this year, Meituan CEO Wang Xing shared his views on AI at a senior management meeting: "AI agents impact me more than ChatGPT. AI is destined to create enormous productivity and will inevitably bring great changes to organizations and working models."
Shortly after that meeting, Meituan held an online meeting company-wide, mainly to promote the installation and use of "Lobster," encouraging everyone to install it and to write as much of their daily work as possible into reusable Skills.
After the meeting, Chen Yujia, who works in merchant operations within Meituan's core local commerce, received a notice to add a section to her weekly report detailing how she used AI to improve efficiency and what Skills could be promoted to the whole team or department. "Then I could feel everyone desperately trying to integrate AI into their work."
One day in April, an Alibaba algorithm engineer received a token consumption ranking for the previous month from his department without warning. He ranked first with 17 billion tokens consumed and was publicly praised. The department head said that annual KPIs and promotion assessments would reference this ranking in the future. But a month later, the new token consumption ranking didn't arrive as scheduled. "Maybe the boss realized this ranking method wasn't reliable."
New rules kept coming. Soon, the department head proposed that employees upload hourly "timesheets" from 11 AM to 6 PM on workdays, with a plugin on the agent automatically recording code and conversation content to generate work summaries—meaning employees couldn't modify their timesheet content. The next day, HR, almost in a confrontational tone, dissuaded the boss from this absurd system.
Incidents like this are no longer surprising. AI anxiety from the top cascades down, with middle managers escalating demands, subtly implying that this is a hidden reporting competition, arms race, and elimination contest.
Although writing Skills wasn't mandatory, Chen Yujia's department head closely monitored each subordinate's token usage, frequently asking about specifics. "He doesn't really know what AI can do, but he says he won't allow anyone in our team to fall behind in this AI wave." Sometimes at after-work dinners, the boss would subtly convey a sense of crisis: "You must use AI, otherwise when the time comes, I won't be able to pull you up."
An engineer working on an Alibaba AI coding product told 36Kr that some business heads had requested their product team to add data tracking points so that "they could clearly see each team member's daily AI usage trajectory."
Some middle managers at Meituan, after receiving layoff targets, even submitted a more aggressive and higher-percentage layoff list—fewer people, higher AI participation, to some extent directly equated to a new-era "management achievement."
AI efficiency became something that any business or function could "tinker with." But a long gap remained between frontline staff and management about what AI can actually do and how to implement it—bosses at all levels have infinite beautiful expectations for AI, while frontline staff try desperately to achieve them but never quite reach that vision, ultimately exhausted by "performing."
Jiang Ling works in customer operations at Alibaba's Taotian Group, responsible for aligning consumer demand with merchant supply. In her view, bosses always "think AI is very intelligent and simple."
Take the common e-commerce anomaly scenario of "order surges." Senior management hopes to proactively identify all "hot products" through full-scale inspection. However, the platform has tens of millions of products a day, far exceeding the capacity of current manpower and tokens. So they can only test on a small scale, selecting a few hundred thousand products, which often results in low hit rates due to the small sample.
"As an employee, you can't refute your boss's expectations, you know?" Jiang Ling said, both agitated and helpless.
Many times, Jiang Ling feels like a donkey with a whip behind her. "Tiredness isn't scary; lack of direction and positive feedback is the scariest. You just keep grinding the mill, not knowing where you're going."
"You can't treat AI like a wishing well," a CTO of an AI company told 36Kr. AI efficiency has many prerequisites. The foundation is data, but many companies haven't done their digitalization well. Moreover, many process bottlenecks lie with "people," which AI alone cannot solve.

"Each generation has its own infrastructure"
For product, operations, and other positions in big tech, the feeling is still one of uncertain anxiety. But programmers have to be the first to accept their fate.
Li Chuan, a frontend engineer at Baidu, was first shocked by AI's capabilities earlier this year when he used Claude Code. "For the same complex requirements, some domestic large models might need five to six rounds of dialogue; Claude does it in two or three rounds and does it better."
His second shock came in April this year. Chinese large model company Zhipu released the GLM-5.1 model. "First, it's cheap; second, its capability is fully comparable to Claude Code."
Li Chuan realized then that his job was at risk. By May, he indeed appeared on the "list."
Like two sides of a coin, one side is that by May 2026, Claude Code's parent company Anthropic had achieved an annualized revenue (ARR) of around $47 billion, growing four to five times in half a year; Zhipu also recently surged to a trillion-dollar market value.
The other side is that the rapid maturity of AI coding has made programmers the hardest hit in this wave of layoffs. "Almost every company targets product and R&D teams first, especially roles like frontend development and test development, which are often seen by bosses as no longer valuable," an HR at an internet company told 36Kr.
In 2025, Li Chuan joined Baidu as a campus hire, becoming a frontend engineer. During the campus recruitment interview a year ago, AI only played the role of a search engine, assisting programming through simple Q&A. The interviewer didn't mention AI at all.
"Frontend" was Li Chuan's ideal job because it's a WYSIWYG (What You See Is What You Get) role—code quality is directly reflected in every detail of the product interface. Every Chinese New Year, telling his family, "Open the Baidu app, that thing on top was made by me," gives him a sense of achievement and "meaning in work."
For years, programmers at big tech were clearly divided into algorithms, frontend, backend, testing, etc. Frontend required higher soft skills like aesthetics and interaction, while backend required more rigorous technical skills. The salary level and "pecking order" in this field were directly tied to "technical content"—frontend was higher than testing but lower than algorithm engineers and backend engineers.
In just one year, everything Li Chuan knew was turned upside down. AI took over large parts of code writing and modification, blurring the boundaries between programmer roles. Even product managers could step into programming.
At an Alibaba development department, in May this year, the department head notified everyone to pause all non-urgent requests. Each team had to develop an agent. For any future business requirements, only product colleagues could directly interface with the agent. Programmers could only modify the agent, not touch the code. The boss also hinted that by October this year, well-performing teams would take over maintaining agents from underperforming teams.
Tencent's CSIG technical team developed a pipeline for fixing bugs in the company's apps—AI fixes the bugs, programmers only need to check after the bug is resolved and click the "confirm" button, and the code is merged. Its current fix accuracy rate is about 50%.
In May, Alibaba established a number of full-stack teams internally, requiring frontend, backend, and test engineers to all become "full-stack engineers" as "super individuals." Starting in June, Meituan has also been promoting the merger of frontend and backend development internally.
Transitioning to "full-stack" is theoretically feasible, but in practice, it's a painful process of peeling away layers.
Han Zhi, suddenly turned into a full-stack engineer, had little time to learn and soon had to start her first "full-stack" project, handling frontend, backend, and testing herself. "Now all my requirements are scheduled backwards, with deadlines set for specific dates," she said. Recently, her work intensity has been maxed out, with tasks still unfinished by 9 PM. "I'm so tired."
But the trend is unstoppable. From late last year to early this year, China's leading companies have been spending as much as possible to encourage programmers to consume tokens, gradually phasing out "ancient programming methods."
At its peak, Tencent CSIG team members enjoyed a token quota of $2,000 per month. As long as the request was reasonable and produced corresponding code output, they could apply to double the quota. Token usage was also included in performance reviews. "When your usage is very low, your leader will ask you why." Therefore, some people would lend their unused token quota to others.
For years, being a programmer at a big tech meant high pay and prestige. They were the foundation of internet companies. The "programmer spirit" meant open source and sharing, concise and elegant code, results-oriented without noise, and excitement when seeing characters dance on the screen.
But times have changed. Almost every programmer interviewed told 36Kr the same thing: "You can't work without AI. If AI 'crashes,' I'd rather spend a lot of time finding a new coding plan than look at code and fix it myself"—talking about "programmer spirit" seems out of place now.
Li Chuan said that the hallmark of a good programmer was learning and iterating, because programming languages had been changing for decades—if you didn't learn, you'd fall behind. He and his friends often went to cafes on weekends to research new technologies. "This group itself is quite competitive." But AI's terrifying iteration speed leaves people speechless.
"If AI coding could stay at the 2025 level, it would level the playing field between people with one or two years of experience and those with seven or eight years, without truly replacing humans, leaving plenty of work outside the 'dialogue box,'" Lin Yue lamented. But technology won't stop for anyone. Now he has no doubt that the extinction of programmers is underway, "like textile workers after the invention of the spinning jenny."

Old growth is gone, new horse races begin
When technology multiplies a company's efficiency, two things inevitably follow—either the same people do more, or the company no longer needs so many people.
"We're not laying off," the CEO of a software company told 36Kr. It took a lot of effort to "train" these programmers who understand the industry and development methods; each is an asset. When AI coding increases programming efficiency by five times, he wouldn't lay off 4/5 of them but would expand the business five times.
That vision is beautiful, but the question is: is there still that much market growth?
Before being laid off, Lin Yue briefly experienced the "liberation" of AI writing code, but soon he became busier. Previously, when the business needed changes to app details, he had to wait for scheduling. Now, requirements pile up faster and faster, regardless of feasibility or importance, with the R&D team told to "just try it out first."
But in Lin Yue's view, many of these requirements are somewhat "mediocre"—the smallest banner ads changing wording details, or floating window ads changing from "free cancellation" to "points deduction." "The product manager tweaks this and that, we do A/B testing, but the results rarely improve."
"The departments with the least growth are the most all in on AI. They always need to find new stories to tell," Cang Shu said. He had worked in both the food delivery business and the drone business. From his personal experience, the former was far more intense in its AI push than the latter.
An Infra engineer who just experienced large-scale layoffs at Meta told 36Kr that after learning to squeeze AI, he and his colleagues wanted to do "a lot of things they didn't have time for before." But now that a large number of people have left, the remaining colleagues have started cutting back on less necessary tasks.
The reality facing everyone is that the star products born in the mobile internet era can no longer substantially drive growth by "doing more work." Some companies are not only not growing but are bleeding badly due to fierce external competition.
In 2025, the food delivery war burned 200 billion RMB among several companies, dragging Meituan's profits and cash flow into a quagmire. This made Meituan, which already had low per-capita profit contributions, the first to enter a layoff cycle. But from another perspective, Meituan's business relies heavily on offline execution, so AI's efficiency potential is relatively small compared to companies with higher online levels. "If even Meituan can reduce staff through AI efficiency, other companies will definitely follow. It's a bellwether," a Meituan employee said.
Baidu, whose traditional cash cow—advertising—is shrinking, and Alibaba's Fliggy and Amap, which have long been marginal and contributed little internally, are in similar situations.
Layoffs in old businesses are unavoidable. Are there opportunities for fresh water?
Some management, when discussing layoffs, tell employees, "The company is also doing AI now. Try to find projects you can work on," a Meituan employee told 36Kr. Recently, Meituan's core local commerce established a new AI Transformation department, mainly exploring the use of AI to streamline internal business processes. Additionally, many core senior and middle managers are personally leading AI-related projects.
Wang Yue, a product manager at ByteDance, told 36Kr that he is working on an internal startup—a B2B AI efficiency tool. "The company encourages such exploration." At the project's inception, they not only voluntarily eliminated the "design" and "testing" functions but also had to emphasize to the review committee how much labor cost this product would save in the future. Another colleague of Wang Yue's is developing an AI customer service agent product, with a 2026 OKR to "help the company lay off xx% of customer service staff."
Today, such projects exist in every major tech company, with a dozen or dozens of small teams each. "Sometimes several teams work on the same direction. Whoever comes out ahead, the company concentrates resources to push them forward." A new horse race has begun.
What's changing, besides business focus, is organizational structure—for example, eliminating more middle managers.
Tencent started implementing a project-based system this year, weakening management ranks and restoring professional ranks for leaders. During its mid-year review, Meituan laid off some L9-level (business unit director) managers and recently completely eliminated the X1 level (previously the lowest management tier), reducing management layers.

Let's say goodbye to the past
Where the AI wave will take people, most still haven't had a "eureka moment."
By mid-June, before the end of his severance buffer period, Lin Yue was aggressively interviewing at Taobao, Kuaishou, and ByteDance. Continuing his "big tech programmer" career was still his preferred path. But these companies' olive branches haven't come yet. "It's too hard," Lin Yue said.
"Finding a job is easy, but once you leave big tech for a mid-sized or small company, you can never return to big tech." In Lin Yue's mind, giving up big tech means a permanent fall; he doesn't want to "settle for second best."
Some have also let go of their "big tech obsession." Three days after leaving Baidu, Li Chuan seamlessly joined a startup. Naturally, his role changed from "frontend engineer" to "full-stack engineer." The startup's main product is an office AI agent, and he even got a raise.
Although everyone says times have changed and programmers' skills are no longer reliable, Li Chuan still has some "technical aspirations," hoping to participate in a product loved by users as a technical professional—and that doesn't necessarily have to happen in big tech.
After leaving Alibaba, Jiang Ling joined an established automobile company. Now her work doesn't have to be forcibly tied to AI. She no longer has to worry daily about "whether the boss's AI tasks can be completed" and, of course, doesn't have to "perform desperately." Jiang Ling's latest project is set to go live on September 30. "These tasks are in my comfort zone, and with plenty of time, I feel so much more at ease."
Recently, every time her department posts a job opening, "a bunch of Alibaba people come to interview, rushing into manufacturing like crazy."
Maybe only 10% of programmers will remain, but Cang Shu doesn't want to look for another big tech job. "I don't want to compete to be part of that desperate 10%."
After being laid off by Meituan in May, he decisively embarked on an entrepreneurial path. Before the AI wave, he had already tried doing something on the side. Back then, just building a community and selling some skills let him taste earning 100,000 RMB a month.
In March and April this year, some of Cang Shu's "students" in the community had already jumped into AI entrepreneurship, "starting their own companies, hiring many people, and here I was working hard for a salary—was that right?" he asked himself.
Now, Cang Shu's startup targets overseas markets, developing systems and standalone products around the needs of rare disease patients. He also shares progress on his Xiaohongshu account "Cang Shu (Quit Monthly Salary Edition)" and overseas social media. Beyond the main product, he is running several small products in parallel to keep his skills sharp. "A small tool takes at most three to four days to complete; a complex system might take half a month." This is far faster than the typical scheduling pace in big tech.
AI might be the most powerful intellectual lever in human history. It can amplify individual capabilities many times, support the launch of most startup products, and allow every good idea to be quickly seen and priced.
Born in 2000, Cang Shu says he was destined to start a business, but without this layoff, he might not have acted now. "The company made the decision for me."
"Don't dwell on the past, forge ahead with passion"—this is the last sentence in Meituan's farewell text to every departing employee, and recently a phrase many big tech employees mention when leaving. In this complex transformation brought by AI, whether leaving big tech or staying, the old paths can no longer be continued.
After a brief "breakdown," it's not about lying down. Whether changing careers or starting a business, those who accept change first might see a different world.