01核心结论Key Findings
以下数字为 Metix AI 数据库口径（数据时点约 2026 年上半年），统计对象 = 当前在职于 Anthropic 的技术人才（工程与研究序列，已剔除社区大使、市场、招聘、产品等非技术角色）。所有数字均为聚合统计，报告不展示任何个人信息。The figures below reflect the Metix AI database scope (data as of roughly the first half of 2026), with the population = technical staff currently employed at Anthropic (engineering and research tracks, excluding non-technical roles such as community ambassadors, marketing, recruiting, and product). All numbers are aggregate statistics; the report shows no individual information.
中位职业经验 13.1 年，72% 有 10 年以上；博士仅 15.5%，技能榜由机器学习、Python、分布式系统、Java 领衔。和 OpenAI 一样，Anthropic 买的是已经把大系统跑起来、并扛过规模的人，而不是论文产出者。Median career experience is 13.1 years, with 72% past the 10-year mark; only 15.5% hold a PhD, and the skills leaderboard is led by machine learning, Python, distributed systems, and Java. Just like OpenAI, Anthropic is buying people who have already stood up large systems and carried them through scale — not paper-producers.
中位在职仅 13 个月（OpenAI 为 17），44% 是过去一年内入职，近四分之三在两年内加入，待满四年的只有 4%。成立于 2021 年的 Anthropic，本质是一支几乎全员仍在第一段股权兑现窗口内的队伍。Median tenure is just 13 months (OpenAI's is 17), 44% joined within the past year, nearly three-quarters came aboard within two years, and only 4% have been there four full years. Having been founded in 2021 Anthropic is essentially a team where almost everyone is still inside their first equity-vesting window.
302 人此前在 Google（占近四分之一），其后是 Meta、微软、亚马逊。与 OpenAI 不同的是，DeepMind（51）与 OpenAI（35）本身就是 Anthropic 可见的人才来源——两大对手实验室合计输送 86 人；Stripe（94）异常靠前，是它独有的基础设施进水管。302 people previously worked at Google (nearly a quarter), followed by Meta, Microsoft, and Amazon. Unlike OpenAI, DeepMind (51) and OpenAI (35) are themselves visible talent sources for Anthropic — the two rival labs supply 86 people combined; Stripe (94) ranks unusually high, an infrastructure feeder unique to Anthropic.
74% 的技术人才共用「Member of Technical Staff」一个头衔，比 OpenAI 的 57% 更极端，连带团队的人也只写作「Member of Technical Staff (Manager)」。想评估 Anthropic 的人，头衔几乎不提供任何信息，只能读履历本身。74% of technical staff share the single “Member of Technical Staff” title — more extreme than OpenAI's 57% — and even people who lead teams are listed only as “Member of Technical Staff (Manager)”. To assess anyone at Anthropic, the title tells you almost nothing; you have to read the track record itself.
02比 OpenAI 还新：四成人过去一年才加入Newer than OpenAI: four in ten joined in just the past year
任期是判断一家公司扩张速度最直接的信号。Anthropic 现任技术人才的中位在职仅Tenure is the most direct signal of how fast a company is expanding. The median current tenure of Anthropic's technical staff is just 13 个月13 months——比 OpenAI 的 17 个月还短。约 44% 入职不到一年，近四分之三在两年内加入，待满四年的只有 4%。这是一支比对手更年轻、仍在高速搭建的队伍。 — shorter still than OpenAI's 17 months. About 44% have been on the job under a year, nearly three-quarters joined within two years, and only 4% have stayed four full years. This is a team younger than its rival and still building at high speed.
03组织很新，人不新：72% 有 10 年以上经验The org is new, the people are not: 72% have over 10 years of experience
和 OpenAI 一样，Anthropic 的组织新、人却资深。中位职业经验Just like OpenAI, Anthropic's org is new but its people are senior. Median career experience is 13.1 年13.1 years，与 OpenAI 的 13 年几乎完全一致；应届与初级是例外而非主体。它把资深经验高度集中，再压进一套比谁都扁平的职级里（见第 07 节）。, almost exactly in line with OpenAI's 13 years; new grads and junior hires are the exception, not the core. It concentrates senior experience heavily, then compresses it into a leveling system flatter than anyone else's (see Section 07).
04技能榜是工程，不是论文The skills leaderboard is engineering, not papers
把 Anthropic 技术人才的硬技能排个序，榜首是机器学习与系统工程——机器学习、Python、分布式系统、Java；深度学习、NLP 这些「研究味」标签排在它们下面，而不是上面。这一点和 OpenAI 一模一样：这是一支为「把大模型与大系统跑进生产环境」而搭的队伍。Rank Anthropic's technical staff by hard skills and the top is machine learning and systems engineering — machine learning, Python, distributed systems, Java; the more “research-flavored” tags like deep learning and NLP sit below them, not above. This is identical to OpenAI: a team built to push large models and large systems into production.
05头号进水管仍是 Google——但对手实验室也在其中The top feeder is still Google — but rival labs are in the mix too
把每个人的过往雇主摊开看，Lay out everyone's prior employers and Google 高居榜首：302 名现任 Anthropic 员工此前在 Google，tops the list by far: 302 current Anthropic employees previously worked at Google, 接近四分之一nearly a quarter；其后是 Meta、微软、亚马逊。真正和 OpenAI 拉开差距的是榜单中段——; followed by Meta, Microsoft, and Amazon. What really sets it apart from OpenAI is the middle of the list — DeepMind（51）与 OpenAI（35）本身就是 Anthropic 的人才来源DeepMind (51) and OpenAI (35) are themselves talent sources for Anthropic，两大对手实验室合计 86 人；而, the two rival labs totaling 86 people; while Stripe（94）Stripe (94)异常靠前，是 Anthropic 独有的金融基础设施进水管。 ranks unusually high — a financial-infrastructure feeder unique to Anthropic.
06博士比例略高，学术管道偏美英而非中国PhD share is slightly higher, and the academic pipeline tilts US/UK rather than China
Anthropic 的博士比例为Anthropic's PhD share is 15.5%，略高于 OpenAI 的 14.4%，但同样不是学术主导——约, slightly above OpenAI's 14.4%, but still not academia-dominated — roughly 85% 的技术人才没有博士学位。从履历中可见的学术机构看，Anthropic 集中在美国顶尖 CS 院校，并带有明显的英国管道（牛津、帝国理工），呼应它在伦敦的第二总部；这与 OpenAI 报告中那条很深的中国本科管道形成对照。 of its technical staff hold no PhD. By the academic institutions visible in their profiles, Anthropic concentrates on top US CS schools, with a pronounced UK pipeline (Oxford, Imperial College) echoing its second headquarters in London — a contrast with the deep Chinese undergraduate pipeline seen in the OpenAI report.
履历中出现最多的学术机构（含读研、博士、博士后及研究经历）：Academic institutions appearing most in profiles (including graduate, PhD, postdoc, and research experience):斯坦福 · 60Stanford · 60哈佛 · 24Harvard · 24MIT · 23CMU · 19UC 伯克利 · 18UC Berkeley · 18康奈尔 · 18Cornell · 18帝国理工 · 9Imperial College · 9牛津 · 7Oxford · 7——一条以美国顶尖 CS 为主、英国为辅的学术管道。 — an academic pipeline led by top US CS schools with the UK as the secondary feed.
07一个头衔统治一切，比 OpenAI 更彻底One title rules them all, more thoroughly than OpenAI
Anthropic 把扁平职级做到了极致。Anthropic has taken flat leveling to the extreme. 74% 的技术人才顶着某种形式的「Member of Technical Staff」——比 OpenAI 的 57% 高出一大截。20 年的系统老兵、刚毕业的博士、甚至带团队的管理者（头衔写作 Member of Technical Staff (Manager)），共用同一行 title。内部当然有层级，但从外部看，这张组织图几乎完全「不可读」。 of its technical staff carry some form of “Member of Technical Staff” — far above OpenAI's 57%. A 20-year systems veteran, a fresh PhD, and even managers who lead teams (titled Member of Technical Staff (Manager)) all share the same line of title. There is of course internal hierarchy, but from the outside the org chart is almost entirely “unreadable”.
其余头衔的长尾The long tail of other titles
Applied AI · 29Research Fellow · 21Software Engineer · 16Research Engineer · 9Research Scientist · 8Engineering · 7Researcher · 6Data Scientist · 5AI Safety Research Fellow · 4除 MTS 之外，是一条由研究类与安全类头衔构成的稀薄长尾——这是 Anthropic 研究与安全底色为数不多的外部痕迹。Beyond MTS lies a thin long tail of research and safety titles — one of the few external traces of Anthropic's research-and-safety DNA.
08这份透视怎么用How to use this X-Ray
如果你在和 Anthropic 抢人If you're competing with Anthropic for talent
主攻入职不足 24 个月的那一层（占全员近四分之三，几乎全员仍在股权兑现窗口内，且比 OpenAI 更早）；从 Google、Meta、微软、亚马逊这些大厂基础设施团队取水，也别忽略 DeepMind 与 OpenAI——Anthropic 自己就在这么挖；别用头衔筛人——74% 都是 MTS，只能读履历底下真正做过的系统。Target the layer under 24 months of tenure (nearly three-quarters of headcount, almost all still inside the equity-vesting window, and earlier than at OpenAI); draw from big-tech infrastructure teams like Google, Meta, Microsoft, and Amazon, and don't overlook DeepMind and OpenAI — Anthropic itself poaches this way; don't screen by title — 74% are MTS, so you can only read the systems they actually built underneath.
如果你想加入这类团队If you want to join a team like this
拿出你真正构建并扛过规模的系统——这里的中位线是 13 年的工程交付，而不是引用数。没有博士也没关系，你属于 84% 的多数。研究与安全方向有少量 Research Fellow 通道，但门槛同样是过硬的工程与研究成果，而非头衔。Bring the systems you genuinely built and carried through scale — the median bar here is 13 years of shipped engineering, not citation counts. No PhD is fine; you're part of the 84% majority. There's a small Research Fellow track on the research-and-safety side, but the bar there is equally about solid engineering and research output, not titles.
这只是一家公司。任何一家，我们都能透视。This is just one company. We can X-Ray any of them.
本报告由 Metix AI 的人才图谱生成——与我们的搜索、匹配产品同一套引擎。想要 Anthropic 的完整名单与可联系的候选人，或者为你正在竞争的某家公司生成同款透视？留个联系方式，我们 1 个工作日内对接。This report is generated from Metix AI's talent graph — the same engine behind our search and matching products. Want Anthropic's full list with contactable candidates, or the same X-Ray for a company you're competing against? Leave your contact details and we'll be in touch within 1 business day.