AI来了。在所有职业中，程序员最先感受到这股风潮的强劲。
从2022年底ChatGPT点燃AI爆发引线到现在，仅仅4年时间，程序员的工作方式、职业路径已经发生了巨大变化。尤其今年AI编程走向真实软件开发场景，编程Agent规模化落地，更是让变化持续加速。
今年“1024程序员节”当日，猎聘发布了一份《AI时代程序员转型趋势报告》，称通过平台上的简历变更情况来看，2026年前三季度有26.4%的程序员样本填写的第一期望岗位与当前岗位方向不同，相较2024年同期高出6.7%。这个数据也被解读为，目前有超过四分之一的程序员正在考虑“换工作”。
近几日，相关话题在程序员社群中也持续引发讨论，不少程序员结合自己的体感和经历给出了来自一线的声音。
热门方向不止AI，“跨界”硬件吸引力增强
程序员的转型与行业需求端的结构性变化息息相关。
在上述报告的统计中，今年前三季度传统程序员岗位新发职位同比下降了8%，而以AI算法为代表的新型技术方向岗位同比增长了20.4%。与此同时，新方向给出的薪资待遇也更加诱人，数据显示，传统程序员平均年薪约29.3万元，而软件与算法岗位的平均年薪达到45.6万元，AI大模型相关岗位的平均年薪更是达到了52.4万元。不少程序员还在讨论中称，真实薪资差距较数据统计更大，传统岗位目前“工资停滞不涨，要求越来越多”，还要面临较大失业压力，一旦能够拿到知名AI公司的技术岗位offer，不仅基础薪资有倍增可能，后续加薪的路径和机会也大得多。不过，从旧入新也并不容易，不少程序员也表示，新方向普遍对程序员基础素质要求更高，针对AI的认知和使用更是有具体考核方式，团队内竞争和淘汰压力往往也较大，“并不是想转就能转”。
实际上，上述报告中转出意向中占比最多的仍是传统岗位，排在首位的“转运维/技术支持”比率为7.2%，排在第二、第三的依次是高端技术职位、项目管理，比率为6.2%、6.1%。人工智能排在第4，比率为6%。
其中，高端技术职位、项目管理属于传统的程序员职业进阶路径，而运维/技术支持岗位，虽然以往也是程序员转岗的常见选择，但正是在AI编程加速替代初级程序员传统的“写代码”工作后，转向那些仍能发挥原有技能和经验积累、但尚未被AI规模化替代的岗位，成为更多人的选择。但也有程序员表示，传统运维的薪资要低于传统开发，很多此类转岗是被动发生的，或是权宜之计，而在AI发展速度越来越快的背景下，这种转岗难免让人有种“往低处走”的感觉。
与此相对的，AI等新技术方向则被认为是“重新选择”的机会。虽然从整体比率来看，换热门新行当的意向目前仍低于在传统路径上转岗、晋升，但从增速来看，前者远高于后者。
据统计，今年前三季度转向AI的转型意向占比较2024年提升2.1%，增幅最大；此外，以往相对属于小众选择的机械设计/制造，同期提升0.8%，半导体/芯片提升0.7%。而且，猎聘的岗位发布数据也显示，硬件端对程序员的需求正处在快速增长期，在今年前三季度面向程序员的新发布岗位中，具身智能与机器人同比增长84.6%，增速最快，其次是半导体与集成电路，同比增长47.8%。而且，相关岗位招聘年薪也普遍在50万元左右，属于高薪去向。随着AI的发展，具身、芯片行业都在着力打造自身的软件团队，提升研发速度并适配AI发展，这也让“跨界硬件”成了程序员转型的新机会。
而与此同时，转型意向中向高端技术岗位、项目管理领域发展的比率反倒比2024年降了2.3%和1.6%。可以说，虽然职业选择的惯性仍在，但程序员“换赛道”的热情正快速超过在传统路径“按部就班”向上走的考量。
从“写代码”到“管理Agent干活”
转型不只发生在不同赛道、各类岗位之间，也出现在程序员每天的日常工作之中。
上述报告也提到，AI能够“写代码”后，程序员职业技能的评价系统面临重塑。尤其对于初级程序员、新人来说，AI能做的事越来越多，自身如何积累判断、排错和独立解决问题的经验成了新问题。统计数据显示，按从业年限划分，从业一年以下程序员的转型意向比率最高，达到39.1%；从业1-3年的程序员相较2024年转型意向增长最快，从17.7%升到了31.3%。以往程序员的“年龄焦虑”通常被锚定在35岁左右，但AI似乎造成了一种“新人焦虑”，越是初入职场，越容易对前景感到迷茫。而以往看来，从业10年以上往往会更多考虑寻找更好的方向，但相比于新人，统计中资深程序员的状态相对较为稳定，今年前三季度从业10年以上程序员的转型意向比率为27.8%，与2024年的数据几无变化。
实际上，就在10月16日，AI模型厂商Anthropic也发布了一份公司内部就业变化的报告。报告显示，在这家以程序员为主的公司内，技术部门初级员工占比已从2025年初的21%降至今年8月的14%，同期初级岗位招聘占比更是从27%降到9%。报告还强调，程序员人均代码提交量在半年内增至原来的5倍。
近两年来，不少AI团队都会提到，代码产能无需再依靠人类程序员“堆量”，AI在其中的贡献越来越大。在前沿团队中，程序员的工作方式也随之变化。
Anthropic的报告提到，今年3月，公司内部Agent的总工时首次超过了人类员工总工时。而程序员的工作重心已经逐渐从“写代码”转向“管理Agent写代码、完成开发工作”。报告称到今年8月，公司内由人类员工发起、引导Agent贡献的总工时达到了人类员工总工时的6倍。报告还称，AI有80%的时间在等待人类输入，看起来，人类反倒成了“瓶颈”。
不少国内AI团队也提出了AI原生（AI Native）的开发模式和组织形态，其中，打散原有开发团队架构，重塑开发流程，让Agent成为贡献代码的主力，而人类程序员转向全员评审，更多承担执行与协作工作，并对目标与结果负责。
虽然目前来看，这些颠覆性的实践更多在AI领域的前沿团队内发生，但AI编程、编程Agent已经在快速向各行各业的开发团队扩散，随着技术的进一步发展，更广泛领域的程序员群体可能都会经历这种变革。
这场程序员的大转型，可能才刚刚开始。（本文首发于钛媒体APP，文｜飞向Tai空）
本文由 AI宝库自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
AI's coming. In all occupations, programmers first felt the strong tide.
The way programmers work and their career paths have changed dramatically in just four years since ChatGPT triggered the AI outbreak at the end of 2022. This year, in particular, AI programming is moving towards real software development, and the programming of Agent ' s size is on the ground, and this is also a continuous acceleration of change.
This year's 1024 Programmer Festival, the hunting of publish, a report on trends in the transition of programmers in the age of AI, stated that, as a result of the change in curriculum vitae on the platform, 26.4 per cent of programmers in the first three quarters of 2026 filled out their first expected posts in a different direction from the current position, which was 6.7 per cent higher than in the same period in 2024. It was also interpreted that over a quarter of programmers were currently considering “replacement of work”.
In recent days, discussions have continued in the programmer community, where a number of programmers have given their first-line voices in connection with their feelings and experiences.
"Transboundary" hardware is more attractive than AI.
The transformation of programmers is closely linked to structural changes at the demand end of the industry.
According to the above-mentioned report, in the first three quarters of the year, the number of new posts for traditional programmers decreased by 8 per cent in comparison with the previous three quarters, while the number of new technical-direction posts represented by the AI algorithm increased by 20.4 per cent over the same period. At the same time, the new directions offer more attractive pay, with data showing that traditional programmers receive an average of approximately 293,000 yuan a year, while the average annual salary for software and algorithm posts is 456,000 yuan, and the average annual salary for AZZ LLM-related posts is 524,000 yuan. A number of programmers have also argued that the real pay gap is greater than the statistics, that traditional jobs are now “stagnating and demanding” and that there is a greater pressure on unemployment, and that once a skilled post at the well-known AI company is available, not only is the base salary more likely to multiply, but the path and opportunities for subsequent salary increases are much greater. However, it is not easy to move from old to new, and a number of programmers have indicated that the new direction generally requires a higher basic quality of programmers, that the perception and use of AI is more specific, and that there is often greater pressure to compete and phase out within the team, “not to turn around”.
In fact, the highest percentage of transfers of intentions in the above-mentioned report is still traditional, with a top “transit dimension/technical support” ratio of 7.2 per cent and second and third, high-end technical positions, project management of 6.2 per cent and 6.1 per cent. Artificial intelligence ranks in 4 at 6 per cent.
Among them, high-end technical positions, project management are part of the traditional career progression path for programmers, while mobility/technical support posts, although previously a common option for programmers, have become more people's choices as AI programming accelerates the replacement of the traditional “writing code” for junior programmers, to those that still build on their previous skills and experience but are not yet replaced by AI. However, programmers have also indicated that traditional mobility is paid less than traditional development, that many of these shifts take place passively or in an ad hoc manner, and that, in the context of the growing pace of AI, such shifts inevitably give the impression of a “lower” approach.
In contrast, the direction of new technologies, such as AI, is considered an opportunity to “reselect”. Although, in terms of the overall ratio, the intention to move to a new line of interest is still lower than that of a move on a traditional path to promotion, the former is much faster than the latter.
According to statistics, the intention to transition to AI in the first three quarters of the year increased by 2.1 per cent compared to 2024; in addition, the design/manufacturing of machines, which had been relatively popular, increased by 0.8 per cent and the semiconductor/chip by 0.7 per cent over the same period. Moreover, the recruitment of publish data also shows that the demand for programmers by the hardware end is growing at a rapid pace, and that, of the new publish posts for programmers in the first three quarters of the year, the number of smarts and robots grew by 84.6 per cent, the fastest, followed by semiconductors and integrated circuits, 47.8 per cent, the same year. Moreover, the annual salary for recruitment to the relevant posts is generally around $500,000, which is a high-paying destination. With the development of AI, the self-contained and chip industries are working to build their own software teams, to increase the speed of R & D and to adapt to AI development, which also makes “cross-border hardware” a new opportunity for programmer transformation.
At the same time, the rate of transition intentions to high-end technical posts and project management declined by 2.3 per cent and 1.6 per cent compared to 2024. It can be said that while the inertia of career choices is still present, the enthusiasm of programmers to “change tracks” is rapidly surpassing the consideration of moving up the traditional path of “scheduled”.
From Writing Code to Managing Agent Jobs
The transition takes place not only on different tracks, but also in the daily work of programmers.
It is also mentioned in the above-mentioned report that the evaluation system of the professional skills of programmers has been re-engineered after AI was able to “writing code”. In particular, for junior programmers and newcomers, AI has been able to do more and more, and its own experience of judgement, error and independent problem-solving has become new. Statistics show that, by length of service, programmers with less than one year of employment have the highest rate of intent for transition, at 39.1 per cent; programmers with one to three years of employment have the fastest growth compared to their transition intentions in 2024, rising from 17.7 per cent to 31.3 per cent. Previously, programmers' “age anxiety” was usually anchored at around 35 years of age, but AI seemed to have created a “new person's anxiety” and the more they entered the workplace, the more confused they were. In the past, it appears that more than 10 years of practice tends to give more thought to finding a better direction, but the situation of senior programmers in statistics is relatively stable compared to new entrants, with a transition intention rate of 27.8 per cent for programmers who have been in practice for more than 10 years in the first three quarters of this year, which is unchanged from the 2024 data.
Indeed, on 16 October, the AI model manufacturer, Anthropic, also publish, submitted a report on changes in employment within the company. The report shows that in this programmer-dominated company, the share of junior employees in the technical sector fell from 21 per cent at the beginning of 2025 to 14 per cent in August of this year, while the share of junior recruitment fell from 27 per cent to 9 per cent during the same period. The report also highlighted that the per capita code submission of programmers had increased five times in six months.
Over the past two years, many AI teams have mentioned that code production does not need to rely on human programmers to “classes”, in which AI is making a growing contribution. In the front team, the way programmers work has changed.
Anthropic reports that, in March of this year, for the first time, the total hours worked by Agent within the company exceeded the total hours worked by human employees. The programmer's focus has gradually shifted from “writing code” to “managing Agent writing code and completing development”. According to the report, by August this year, the total number of hours worked by Agent, initiated by human employees, was six times that of human workers. The report also states that AI has been waiting 80 per cent of its time for human input and that it appears that human beings have become “bottlenecks”.
A number of national AI teams have also proposed a development and organizational model for AI Native, in which the original development team structure is broken down, the development process is reshaped and Agent is the main contributor to the code, while human programmers turn to full-time reviews, more executive and collaborative work, and more responsible for objectives and results.
While these subversive practices are now more present in the front line of the AI field, AI programming and programming Agent has spread rapidly to development teams across all sectors, and this change may be experienced by a wider group of programmers as technology further develops.
This major transformation of programmers may have just begun. (This article was first posted on titanium media app.
Auto-fetched and translated by AI Gem Box. All rights belong to the original author.