Wink Pings

AI Agent Trends Are Hidden in Big Tech Job Descriptions

Looking at recruiting JDs cannot replace judgment, but it can tell you what capabilities companies are willing to pay for. The four screenshots in the original post and the counterarguments in the replies are worth examining together.

Judging where AI Agents will go in the next two years has a very low-cost method: look through big tech job descriptions. JD stands for job description, the task list that recruiters write for candidates. The original poster, MX狗哥, said that AI Agent job descriptions at companies like Xiaohongshu, ByteDance, Alibaba, and OpenAI are more useful than reading a hundred industry analyses. The original post included four screenshots.

Job description screenshot 1/4 attached to the original post

Job description screenshot 2/4 attached to the original post

Job description screenshot 3/4 attached to the original post

Job description screenshot 4/4 attached to the original post

Screenshots do not explain trends for you, but they can expose a question: what work are companies preparing to have Agents do, and what capabilities are they willing to pay for? A commenter said in a reply that a job description will not tell you the truth, but it will tell you what capabilities a company is willing to pay for. That sentence is more direct than many industry reports. Reports can write far into the future; JDs cannot, because they need to hire people to do the work.

But do not treat JDs as a crystal ball. Some commenters reminded that big tech JDs lag behind, and a small number of startups may be further ahead. Another commenter said that looking at a single JD is like looking at a wish list; tracking changes in the same team over several months provides more information: which requirements remain, and which buzzwords appear once and then disappear. This observation is closer to a method. A trend is more like a certain requirement repeatedly appearing, appearing across teams, and beginning to be written into specific responsibilities.

For people with a development background, the tech stack and task descriptions in a JD can be viewed as demand samples. For ordinary users, they can at least understand what troubles AI Agents need to solve when being deployed: tool calling, workflow integration, evaluation, permissions, and cost. I will not expand here by guessing the contents of the screenshots; the four images in the original post already provide samples. What should be noted is that buzzwords in recruiting requirements are not the same as the actual work of the position. Buzzwords attract people; responsibilities screen people. Looking at responsibilities is more reliable than looking at titles.

There is also cold water. A commenter said that after reading JDs and then submitting resumes, the education section alone filters out half the people. Trends and opportunities are not the same thing. Knowing where a company is heading does not mean the door will get wider. There is no need to avoid this contradiction. Another commenter said that JDs are an observation angle and should be read together with products and technical documentation to make more reliable judgments. Another reply was more practical: someone uses AI to make weekly recruiting reports for positions, aggregating positions to see changes. That is more useful than saving a bunch of screenshots.

If you really want to use JDs as a signal, do not do three things: do not look at only one company, do not look only once, and do not turn every requirement into your own to-do list. Choose three to five target companies, look every two or three weeks, record requirements that are added and those that disappear, and then see whether their products have corresponding moves. If you keep watching for several months, there will be much less noise. Big tech JDs lag, but that lag also has value: it can tell you which capabilities have already moved from experimentation into the recruiting process.

Original post link: MX狗哥's tweet

发布时间: 2026-11-18 13:31