AI Coding's Impact on Programmer Employment: Outsourcing Talent Pools Shrink, Hiring Shifts to Assessing AI Tool Invocation

The main body of the article

Programmers were once one of the most enviable jobs in the internet industry.

A big-tech résumé, a million-yuan annual salary, and technology changing the world—these are the most dazzling labels of this profession. Young people crowded into big tech companies, and companies competed for technical talent. Mastering a programming language often meant a respectable income and a relatively certain career prospect.

The halo over programmers is fading.

The impact is first reflected in income.

"The situation is not good—there are a large number of programmers waiting for employment." Wang Yonghong, deputy general manager of Beijing Nuanyi Suixing Health Management Co., Ltd., told a Times Weekly reporter. He has been engaged in program development for many years and is pessimistic about programmers' employment prospects: "It will probably be very difficult for salary levels to return to the past."

Falling income is only the surface. What worries and troubles the programmer community most is that—under the AI wave, code generation tools have taken over basic writing and debugging, and demand for entry-level positions has plummeted. The industry is undergoing structural reshaping, and the core value of programmers is accelerating its shift from "hand-writing code" to "architecture design and AI collaboration." Do companies still need that many programmers, and which programmers can match companies' real needs?

The World Economic Forum's "Future of Jobs Report 2025" predicts that by 2030, 92 million jobs will be displaced globally, while 170 million new jobs will be created. A "Skills Trends Report in the AI Era" released by Tsinghua University shows that the AI replacement rate for entry-level programmer positions has reached 87%.

"It's not that AI has replaced people, but that after job boundaries dissolved, your upstream and downstream simply did your work for you," a senior engineer, "Hailalu Programming Guest," told a Times Weekly reporter. In his view, the biggest impact of AI Coding on big tech R&D systems is not fully automated development, but the crossover of roles and merging of capabilities in daily R&D.

In the past, software development at big tech companies was like an assembly line with a clear division of labor: frontend, backend, and testing each performed their own duties, with clear technical boundaries between positions. Today, AI Coding is weakening these boundaries. Backend engineers can use AI to complete frontend development, and frontend engineers can also use AI and full-stack frameworks to write database logic and independently complete backend development for small and medium-sized businesses.

Jobs have not disappeared immediately, but the scope of work within a single job is expanding. Hailalu Programming Guest believes that AI gives engineers with a global perspective comprehensive delivery capability across frontend and backend, testing, and even product links. When one person can use AI to complete work that previously required collaboration among multiple people, companies' demand for single-function positions will naturally be affected.

This is the more hidden but potentially more far-reaching impact of AI Coding on the programmer job market: it may not directly replace a certain programmer, but it may lead companies to discover that completing the same work no longer requires as many people as before.

Claude Code, Cursor, Codex, Trae, Pieces... Over the past two years, AI Coding tools have emerged intensively. These tools were initially intended to help programmers improve efficiency, but now they are beginning to change programmers' ways of working in turn, and even redefine how many people a development team needs.

The core value of AI Coding is not just helping programmers complete a few lines of code, but enabling AI to gradually possess the ability to understand requirements, call tools, write code, run tests, and fix errors. Especially after the emergence of Coding Agents, AI is no longer just a code assistant waiting for instructions, but an agent that can continuously execute multiple steps around a development task.

This means programmers no longer need to personally complete every line of code; instead, they can hand some development tasks to AI while they are responsible for breaking down requirements, setting goals, checking results, and handling exceptions. Work that previously required collaboration among several people may now be completed by one engineer skilled in using AI.

What is being impacted is not only programmers' workload, but also the division-of-labor system that formed around software development in the past.

Hailalu Programming Guest told a Times Weekly reporter that single job categories and pools of outsourced personnel are shrinking sharply. Taking non-core business lines as an example, companies in the past might have maintained a dozen or so outsourced and junior developers year-round, responsible for routine campaign pages, mid- and back-office forms, and other work. Now, these tasks can be handed to Agents for execution, with final review by a small number of engineers.

Feng Xu, general manager of Shanghai Renruan Technology Co., Ltd., also provided a specific figure to a Times Weekly reporter: after using Coding Agents, the company's per-capita number of new code lines rose significantly, with actual testing showing an increase of about 40% in per-capita new code volume.

The speed of code output does not equal a proportional increase in software delivery efficiency.

Feng Xu said that the coding speed for individual tasks has indeed become faster, but the end-to-end development cycle has not shortened proportionally. At the same time, waiting time for PR (pull request) review has instead increased, and the burden of rework and review has also risen somewhat.

This reveals another side of the AI Coding efficiency revolution: code generation is becoming cheaper and cheaper, but the costs of verifying code, maintaining architecture, handling business logic, and ensuring system stability have not disappeared in sync. AI pushes the bottleneck in the production link backward, and code review, system integration, and quality control are becoming new efficiency constraints.

But what is more noteworthy than changes in programmer positions is the possible contraction of management layers that may follow.

Hailalu Programming Guest believes that the core work of some middle-management positions in the past was to break down requirements, arrange schedules, coordinate front-end and back-end tasks, and then summarize progress and weekly reports. But when three engineers with full-stack capabilities work with AI and can complete the work that previously required eight people, this management system that relies on multi-person division of labor and layer-by-layer coordination will also be impacted.

On October 27, Microsoft launched a voluntary departure plan involving about 60,000 employees. As of the end of June this year, Microsoft had about 223,000 full-time employees, and the number involved in the plan accounted for about 27% of the total workforce. Eligible employees need to decide whether to accept the plan before early December and leave the company by the end of this year.

This plan cannot directly prove that Microsoft is laying off employees because of AI replacement, nor can it be simply attributed to a result spawned by AI. But against the backdrop of AI continuously improving R&D efficiency, it has become a trend worth watching that tech giants are re-examining organizational scale, workforce allocation, and management layers.

The hiring wind is shifting

What AI Coding is changing is not only the way programmers work, but also the standards companies use to hire programmers.

In the past, when companies hired programmers, they valued the basics most: whether programming languages were proficient, whether code was written fast enough, and whether development tasks could be completed independently. After all, software development is a labor-intensive job, and companies need enough engineers to turn requirements into lines of code.

But now, code seems to be worth less and less. Feng Xu told a Time Weekly reporter that two or three years ago, the company mainly valued three abilities when hiring programmers: fast manual coding, proficient syntax, and the ability to quickly pile up features. Now, as AI Coding tools mature, this set of standards is changing.

Feng Xu said that the company now values more whether programmers can clearly break down vague requirements, whether they can use AI to accelerate development, whether they can review code and find problems, and whether they can complete architecture design and business closed loops. "The ability to write code is the bottom line. Acceptance, debugging, system design and judgment, and cross-end delivery are what our company needs now."

Feng Xu also revealed that the company's written test currently mainly examines three parts: Java, Python, and the ability to invoke AI tools.

Written test questions provided to the reporter by Feng Xu

"In the past, tests loved to ask about memorizing framework source code and hand-writing obscure algorithms; now these can be generated in a second. What companies value now is how you divide system boundaries and design data flows when facing a vague business request," said Hailalu Programming Guest. In the future, what determines a programmer's salary ceiling is not how many lines of code they can write in an hour, but how many lines of AI-generated code they can review with high quality and backstop in an hour.

Hiring plans have also begun to adjust accordingly. It is reported that many software development teams are increasing the introduction of college student interns while reducing social recruitment quotas for some basic positions. Interns have lower employment costs, higher work enthusiasm, and relatively lower management costs.

Stripe recently posted summer 2027 software engineering high school internships, open to students who have not yet graduated from high school, requiring interns to participate in real software engineering projects, write and test code, and undergo technical review by the engineering team. As AI lowers some of the barriers to software development, companies are starting to look earlier for people with technical potential who can use AI to complete tasks.

The deeper impact lies in the lowering of the software development threshold and the expansion of programmers' competitive boundaries. This is an easily overlooked change.

In the past, programming ability was a natural threshold, and people who did not know how to write code found it difficult to develop software independently. Often, product managers proposed requirements, designers were responsible for the product interface, and programmers were responsible for turning requirements into code.

Now AI Coding is lowering this threshold. With tools such as Cursor and Claude Code, even people without a professional programming background can describe requirements in natural language and let AI generate code, build pages, and develop simple applications.

A recent survey released by Zapier shows that among workplace professionals who use AI tools to develop and deliver software, 34% do not have a formal programming background. Among them, more than half have developed tools that are already aimed at customers or the public, and 54% said the tools they developed are still in continuous use.

A research paper published by IEEE in 2025 also summarized this trend as Citizen Development. Generative AI has further lowered the threshold for software development, enabling business personnel who originally did not have professional programming capabilities to participate in application development.

In the future, the competitors programmers face may no longer be only other programmers. When product managers can build prototypes themselves, designers can directly implement interactive pages, operations staff can develop data analysis tools, and business departments may no longer need to hand all software requirements to the technical department.

Programmers' new value

In the face of AI Coding, where is the way out for programmers?

Senior programmer Haisong (pseudonym) told a Time Weekly reporter: "Focusing only on technology itself is no longer useful; the greater need is to move toward the front office."

This change has already appeared in leading tech companies.

Google CEO Pichai once revealed that 75% of new code inside the company has already been generated by AI. Programmers' focus is shifting from writing code personally to reviewing code, breaking down requirements, and solving complex problems.

In the future, companies may care more about three questions: whether they can translate business needs into technical solutions, whether they can harness AI to complete complex tasks, and whether they can take responsibility for the final delivered results. The professional boundaries of programmers may therefore also extend toward product managers, technical architects, and business leaders.

A signal worth noting is that, just as programmers begin to worry about their jobs being replaced by AI, a new role—FDE (Forward Deployed Engineer)—is emerging at tech companies.

A 2026 workforce report released by LinkedIn shows that from 2023 to 2025, global FDE job postings grew 42-fold. A Time Weekly reporter found that major Chinese internet companies offer FDE experts monthly salaries of 30,000 to 80,000 yuan, and U.S. AI companies such as OpenAI and Anthropic also offer relatively high pay for related positions.

The emergence of FDE reflects, to some extent, changes in the AI industry's demand for technical talent.

Traditional software engineers usually develop standardized products inside the company, and then sales, implementation, or customer success teams deliver the products to customers. FDEs, by contrast, need to go directly to customer sites, understand real business processes, identify technical problems, and work with customers to develop and deploy solutions.

Take a manufacturing company that hopes to use AI to reduce equipment failure rates as an example. Traditional software engineers might be responsible for developing a predictive maintenance system, while implementation consultants handle deployment and training. An FDE, however, needs to understand factory equipment, production processes, enterprise data systems, and AI model capabilities at the same time, determine which links are suitable for introducing AI, and then participate in agent development, API integration, data access, workflow transformation, and actual deployment.

Haisong believes that in the future, versatile talent will be more favored by companies, and relying solely on technical ability is no longer enough.

Hailalu Programming Guest also believes that programmers should not limit their sights to "keeping the rice bowl of writing code." When programmers shift from being people who write code to people who design and orchestrate agent workflows, their career stage may expand from narrow IT departments to broader industry business scenarios.

In his view, fields such as finance, legal affairs, content production, enterprise operations, and cross-border e-commerce may also gradually introduce Multi-Agent workflows. Different agents undertake different tasks and, through task orchestration, contextual knowledge bases, and quality verification mechanisms, complete complex work that previously required collaboration among multiple people.

English text is being prepared.

(Source: Time Weekly) Mid-body embedded advertisement