AI Hiring Brief · 2026-11-09
32 frontier AI labs · 6 roles added in 24h · agihunt.info/en/jobs
Today at a glance
- Quiet day: Across the 32 companies tracked, only 6 roles were added in 24 hours, and the 5 companies that added anything did so in ones and threes. Closures (95) far outnumbered additions, and most of the closures came from one company.
- Physical build-out at OpenAI: The one new OpenAI role is "Civil & Structural Architect, Power Generation", in the Scaling department and listed as US - Remote. It is another power and construction title, not a software one.
- Cybersecurity RL at Anthropic: Anthropic's one addition is "Research Engineer, Cybersecurity RL", in AI Research & Engineering, based in San Francisco, CA or New York City, NY.
- Low-level infrastructure at SpaceXAI: SpaceXAI added "Lead Engineer, AI Infrastructure (C++, C)" in Engineering, in Palo Alto, CA and Seattle, WA.
- Product roles at Baidu: Baidu AI (百度 AI) added 3 roles, the only China-based additions of the day. They are product and solution positions, not research roles.
Company moves
OpenAI added 1 and closed 5, leaving 816 open. The new role, "Civil & Structural Architect, Power Generation", is a construction and energy title in the Scaling department. It points to power supply being treated as an engineering problem the company staffs for directly.
Anthropic added 1 and closed 1, with 675 open. "Research Engineer, Cybersecurity RL" combines security with reinforcement learning. It is a research-engineering seat, not a go-to-market one.
SpaceXAI added 1 and closed none, with 304 open. "Lead Engineer, AI Infrastructure (C++, C)" is a lead-level infrastructure role, with the language requirement stated in the title.
Baidu AI (百度 AI) added 3 and closed 2, with 478 open. The titles are "短漫剧AI解决方案(J107169)" (short-drama AI solutions) in MEG, based in Beijing and Shanghai. The other two are "公共出行产品经理(J106938)" (public mobility product manager) in IDG, in Beijing, and "AI 产品专家(Agent 知识引擎方向)(J107146)" (AI product expert, agent knowledge engine track) in PSIG, in Beijing. All three are product or solution roles, and one of them is aimed at agents.
Closures without additions: ByteDance Doubao/Seed (字节跳动豆包/Seed) closed 74 roles in one day and added none, leaving 775 open. Its 7-day additions were 10. StepFun (阶跃星辰) closed 8 and added none, leaving 330 open. NVIDIA closed 3 and added none, leaving 585 open. Meta AI, Meituan LongCat (美团·龙猫) and Google DeepMind had 1 closure or fewer each. The data does not say why any of these roles closed.
The other companies in the data showed no change in the last 24 hours.
What it signals
The 7-day figures give a better picture than a single quiet day. Over the past week, the 32 companies added 419 roles: 344 at the 19 global companies and 75 at the 13 China-based ones. The global total is concentrated at the top. OpenAI added 70, Anthropic 61, Meta AI 49, NVIDIA 42, SpaceXAI 34 and Google DeepMind 23. Apple AIML added 21. Those seven companies account for 300 of the 344 global additions. Microsoft AI (622 open) and Mistral AI (168 open) added nothing in the week, so a large open count does not by itself mean current hiring.
The China-based labs move differently. StepFun is the exception, with 33 additions over 7 days and 330 open. Baidu AI added 8 and ByteDance Doubao/Seed added 10. Most of the rest added between 0 and 5: Moonshot AI (月之暗面 Kimi) 3, DeepSeek (深度求索) 2, ModelBest (面壁智能) 5, Tencent Hunyuan (腾讯混元) 2, MiniMax (Hailuo) 4, 01.AI (Yi) 5, SenseTime SenseNova (商汤日日新) 1. Zhipu AI (智谱 AI) and Alibaba Tongyi Qwen (阿里通义千问) added none. ByteDance is the clearest case where the 1-day and 7-day views differ: 10 added in the week, then 74 closed in a single day. That looks like a clean-up of listings, though the data alone cannot confirm it.
The three Baidu titles all sit on the product side, and one explicitly names agents. That fits a pattern where the China-based companies posting new roles skew toward productization, while the global leaders that are adding roles at volume post a mix that includes infrastructure and physical build-out. Today's OpenAI and SpaceXAI titles are examples of that infrastructure end.
One day of data is thin, and the 7-day totals are a more reliable guide to direction. For now the supported reading is that hiring is heavily concentrated at a handful of global labs. Most other companies, global and Chinese, are adding roles slowly or not at all.