Under the impact of AI programming, programmer positions are polarizing; practitioners are turning to traditional industries and one-person companies

In 2026, the capabilities of AI programming tools improved dramatically. Benchmark figures such as DHH, the father of Rails, publicly shifted toward AI programming. The global tech industry accelerated layoffs because of AI, programmer positions have become polarized, and practitioners are turning to traditional industries or starting their own businesses to find a way out.

**Key Points**

**1. AI programming has become mainstream in the industry, and benchmark figures have publicly shifted**

DHH, the father of Rails, previously insisted on writing code by hand and criticized AI, but in September 2026 at the Rails World conference he announced that the company would no longer use handwritten code as its regular development method. Nearly 75% of new code inside Google is now generated by AI, compared with only 15% two years ago.

**2. The global tech industry is accelerating layoffs because of AI, and the job market is clearly shrinking**

In the first half of 2026, cumulative layoffs in the global tech industry exceeded 120,000, already matching the full-year scale of 2025, most of them related to AI. Big companies are indirectly reducing positions through harsh KPIs and replacing full-time staff with outsourcing. The weight of technical value is declining, while the weight of business understanding is rising.

**3. Programmer positions are polarized, and talent in the middle is being squeezed**

Companies are only willing to pay for deeply experienced experts or potential talent from prestigious universities, while the value of ordinary programmers in the middle is being diluted by AI. High-end talent in algorithm positions is in demand, the low-end market is shrinking, and outsourcing and basic positions have become the hardest-hit areas of AI replacement. Alibaba P7-level Java programmers can no longer secure headhunter budgets.

**4. Programmer transition paths: traditional industries and one-person companies**

Until early 2026, DHH (David Heinemeier Hansson) was still unwilling to give up the "ancient craft" of writing code by hand.

This programmer, known for his love of programming, created Ruby on Rails, a software development framework that has helped countless people turn ideas into websites and businesses. For more than twenty years, he always believed that writing code itself is worth loving. Facing the temptation to let AI do the work, he once wrote in a blog that he would rather retire than permanently hand the keyboard to AI.

But on September 23 this year, standing at the Rails World conference in Austin, USA, he said the company no longer treats handwritten code as the regular way to develop software, and announced to peers, "Pencils down." This benchmark figure who long criticized AI code and defended human programmers also had to compromise with technology.

DHH believes programmers will not completely lose their value. But he also admits that some companies may no longer need so many programmers. Long-term progress beneficial to all humanity may bring short-term difficulties to specific people.

At its Cloud Next conference in April, Google disclosed that nearly 75% of new code inside the company was already generated by AI, compared with only 15% two years ago. Statistics from Layoffs.fyi show that in the first half of 2026, cumulative layoffs in the global tech industry for the year exceeded 120,000, already matching the full-year scale of 2025, most of which were related to AI.

More programmers have directly felt this during interviews. Many of them may have spent their entire careers just barely learning everything AI can do, but now these abilities are no longer valued. Companies hiring them have adopted a clearer attitude—"you'd better know something AI doesn't."

Acceptance

Wu Tianyu began using AI programming in 2023. At that time, Vibe Coding products represented by Cursor were not yet mature, so he only used ChatGPT to write code and then copied and pasted it. By early 2024, he clearly felt that the code AI wrote was better than his—more concise and faster, "able to finish a day's work for me in a few minutes."

At that time, Wu Tianyu was in the second year of his PhD, majoring in "Intelligent Manufacturing and Robotics." The senior students in his lab generally had very high programming skills and almost all wrote code by hand. When he occasionally opened the AI programming window, he had to secretly avoid his classmates' eyes.

What followed were exactly the two years in which the programming capabilities of AI large models grew wildly. From early 2024 to mid-2026, the flagship models of six companies—OpenAI, Anthropic, Google, Zhipu AI, Minimax, and Kimi—showed major improvements in mainstream programming capability tests such as HumanEval, MBPP, and SWE-bench Verified (including the first generation), with the first two tests generally rising by nearly ten percentage points.

Wang Guangyan, who does back-end development at a major internet company, was also extremely distrustful of AI at first, because early models still had serious hallucinations. "At the time I thought it might be like AlphaGo—flare up in a leap, and then development would slow down." It was not until the past year of intensive use that he found the efficiency gains of AI Coding increasingly obvious.

In the past, Wang Guangyan might spend 60%-70% of a day writing code, and the rest of the time confirming requirements, doing design, and investigating problems. After AI entered the workflow, this time allocation ratio reversed.

Wu Tianyu also "gave up resistance" in mid-2024 when he started his graduation project. His advisor wanted him to independently develop a robot-related app, and because the workload was huge, he had to rely on AI.

After AI easily produced some UI interfaces that he could not have implemented, Wu Tianyu's anxiety and resistance all dissipated. He realized clearly that what AI replaces is the mechanical thinking part of "ancient programming," and instead gives people more time to think about innovation points and aesthetic details. He not only began using it openly, but also actively recommended it to his senior classmates.

Layoff data caused by AI in the global tech industry. Image source: layoffs

More interviewees said this is indeed unfriendly to workers. Many of them believe that 2026 is the "first year" of layoffs in the tech industry, and that within the next 1-3 years, the industry will lose 30%-50% of positions, with the final reduction possibly even higher—perhaps 80%-90%, or perhaps it will cap at some number.

For example, one rather counterintuitive judgment is that the career impact of AI programming on programmers over 35 may be positive.

Before AI emerged, the saying that "programmers can only work until 35" had already become a popular consensus. For these "older programmers," the career cycle itself is only so long; whether AI exists or not will not cause too much additional shock to their careers.

"Indirect replacement"

Most of the time, technological replacement does not happen suddenly.

At big companies, it usually manifests in an indirect way. For example, setting harsh KPIs for teams to promote internal competition and conduct survival-of-the-fittest selection in the form of horse racing, so that the team headcount can be kept at 70%-80%, or even half, of what it was in the past.

Another more obvious change is that interview experience or techniques oriented toward "ancient programming" have become ineffective.

A fresh graduate who had just gone through the spring recruitment said that last year's experience no longer applies this year, and every specific change requires him to figure it out through his own struggles. "If you are not particularly matched with the requirements of this position, he will ask you some very difficult questions to prove that you have the potential to solve unknown problems."

According to job seekers' summaries of their experience, companies now only recruit in two ways. One is to spend money buying experience: "You have deep experience in this field and have been through enough pits and stepped on enough mines." The other is to spend money buying potential, such as fresh graduates from Tsinghua and Peking University, or PhDs from Ivy League schools. For talent in the middle zone, companies' willingness to pay is declining, "because AI can do most of the work."

Headhunters are also very sensitive to changes in the industry's "water temperature."

What he observed is that big companies' requirements for vertical business expertise in talent have risen sharply, causing the range of candidates headhunters can choose from to shrink dramatically. An Alibaba P7-level Java programmer was "very easy to sell" seven or eight years ago, but today there is no headhunting budget for them at all, and almost all of it has shifted to RPO (mass human resources outsourcing) or ITO (project technology outsourcing). What is still in demand now are high-end vertical positions at at least the architect level or P8 and above.

Behind this is the decline in the weight of technical value and the rise in the weight of business understanding.

Image source: Maimai "2026 Spring Recruitment Workplace Insight Report"

While the headcount of regular frontline employees has shrunk sharply, many big companies and even second-tier internet companies are launching outsourcing on a large scale to fill basic, highly replaceable work, such as testing, front-end, and data annotation.

Not long ago, news that an internet giant was massively cutting outsourced workers spread widely. "I dare not say the exact numbers, but the large-scale cutting of outsourcing is something that truly happened," Qiu Yuewen said. "The essential reason is AI replacement."

Looking across all programmer positions, algorithm roles have the weakest replaceability because they are highly dependent on "understanding the business and issuing algorithm strategies." But even in algorithm roles, there is a polarization in which high-end talent is in demand while the low-end market shrinks.

Change the way out, change the mindset

In the past two years, promoted by Musk and others, a futurological hypothesis has become increasingly popular among the public: that carbon-based humans may only serve as a transitional carrier of intelligent civilization, with the essential purpose of creating artificial intelligence and migrating intelligence to silicon chip carriers in order to complete civilizational iteration.

There is no shortage of people who really face the future with this mindset. But they are not anxious or panicked; they merely take rational risk precautions, such as reducing meaningless consumption and investment and maintaining cash flow.

But on the positive side, under an irreversible trend, do programmers have a Plan B for their professional lives?

At the headhunting company where Qiu Yuewen works, one business direction with considerable growth is sending departing programmers from top-tier big companies to traditional industries such as pharmaceuticals, retail, fast-moving consumer goods, and manufacturing. These industries are facing digital and intelligent transformation and have a large number of internal AI Agent development needs.

"Now many programmers who have AI skills but do not want to continue involution in the internet industry choose to go to traditional industries to lead AI." Their total salary packages will shrink, but because the work intensity is far lower than in the internet industry, their real hourly pay has not changed much.

Traditional enterprises have another advantage: they are less strict about age than the internet industry. "Even 40-year-olds can get in, and if the level is a bit higher, those under 45 are also acceptable." The main R&D staff at some pharmaceutical companies are mostly PhDs, and their average age is naturally higher than in the internet industry. In addition, these enterprises relatively recognize the technical foundations of big companies and strongly favor people with internet backgrounds transferring over.

This can be regarded as a "forced landing path" to traditional industries, which can at least extend a career by another 5 to 10 years.

In addition, AI programming has also opened a new route of "one-person company (OPC)" for many people.

Zhou Ziyou, who majored in computer science as an undergraduate, is currently overseas, working as a senior developer at a big company while starting a small company of his own.

Outside his full-time job, his daily personal business is to receive tasks or requirements from clients and then write "Tickets" for AI Agents. Because the Agent he has trained over time has already mastered a large amount of background knowledge, Zhou Ziyou usually only needs to give it basic information (sometimes even just the chat records with the client). After programming is launched in the background, he only needs to conduct code review and be responsible for final quality control and branch merging.

"Previously our (startup) team had always been two people. In early 2025 I felt we had to hire, but for the whole year we did not hire a single new person, and the team's delivery volume was still several times that of previous years," Zhou Ziyou said. "I am confident I can compare efficiency with a traditional 10-20 person programming team that does not use AI tools, and my quality is even better."

When Wu Tianyu was an undergraduate, his wish was still to become a big-company programmer, but he has now given up that idea and chosen to start a business, running a robotics company in the embodied intelligence field.

In Longhua District near Shenzhen North Station, Wu Tianyu's team rented a 180-square-meter residential apartment. Of the four bedrooms, three are used as offices and one as a warehouse. Wu Tianyu's workstation is in the living room, which is also the robot "testing ground," with four prototype robots and one robotic arm placed there.

It is 7:30 p.m. on a weekday, and later a technical executive from an embodied intelligence unicorn will visit to talk with him about cooperation in human-computer interaction; early tomorrow morning, he will also meet an investor.

"Full of challenges." That is how Wu Tianyu describes his current life. "I do not resist it, nor do I fear it; I even enjoy it."

(At the request of the interviewees, Wang Guangyan, Qiu Yuewen, and Zhou Ziyou are pseudonyms in this article.) AI industry signal channel: Frontier Technology. This content is authorized by the author for publication, and the views represent only the author and not Huxiu's position. If you have objections or complaints about this manuscript, please contact tougao@huxiu.com. Everyone who is changing and wants to change the world is in the Huxiu APP[ ]]]