What Skills Do Anthropic Employees Actually List?

Ask most people to guess the most common self-listed skill among Anthropic employees, and "Leadership" is not the first answer. Ours was not either, and we only found out because we could actually check: a random sample of 1,000 current-employee profiles, drawn straight from our own indexed dataset rather than a survey or a press release.

780 of the 1,000 profiles in the sample list at least one skill, averaging 27 skills each. Below is what the top 20 actually look like across the sample, and the one thing about how a self-listed skills field works that explains why the list looks the way it does.

Counted across the sample: how many of the 1,000 profiles list each skill at all, not how prominently, since the source profiles do not expose a reliable ranking within a person's own skill list.

The honest explanation is not that Anthropic is quietly a management-heavy company. It is that the self-listed skills field on a professional profile is cumulative across an entire career, not scoped to a person's current role. Almost nobody prunes old skills when they change jobs, so a profile still lists "Customer Service" from a retail job a decade ago, "Public Speaking" from a debate club, or "Microsoft Excel" from literally any office job that ever existed.

That is exactly why generic, career-spanning skills (Leadership, Management, Microsoft Office, Public Speaking) dominate the top of the list, and why a technical skill like Machine Learning sits at position 18 with 119 mentions, well behind Java, Python, and SQL. It is not that fewer than 12% of a frontier AI lab's staff know machine learning. It is that "skills" here means "every skill this person has ever self-tagged," and a research org still employs plenty of recruiters, salespeople, and operations staff whose own skill lists were never going to say Machine Learning in the first place.

This is real, queried data, not an estimate: a random 1,000-profile draw from our own indexed dataset, executed against the live table. What it cannot do is stand in for role-specific hiring data. A skills frequency count across an entire company, including every function in it, answers "what does this company's workforce look like on paper," not "what does Anthropic look for in an ML engineer."

Getting that second, narrower question right requires filtering by current title before counting skills, comparing an "Engineer"-titled subset against the company-wide baseline above. That is a real follow-up, and the gap between the two numbers would itself be the finding.

The same person and company data this API serves is what this sample was drawn from: a random selection scoped to profiles where `current_company_id` matches Anthropic, then every self-listed skill on those profiles counted once per profile.