Read the Conceptual Reasoning Fellow Posting Before You Apply
A posting tells you what a role needs, and it often tells you more about who the role is for than a job title does. Anthropic's Conceptual Reasoning Fellow posting is short, and one of its duties is a skill you can practice before you apply. We read it on October 15, 2026.
The dates. The posting says applications close October 28, 2026, and that offers are planned by November 20. The fellowship is a fixed-term contract from January 12 to May 8, 2027, about four months and full-time. Fellows are engaged through an employer of record, not directly by Anthropic, and the posting says some fellowships may be considered for an extension.
The pay and place. It lists 100,000 to 167,000 USD for the four-month fellowship, which it also writes as 300,000 to 500,000 USD annualized, and other benefits. The default location is San Francisco, and remote work in the US is considered in certain circumstances. The posting asks all Fellows to attend a one-week in-person orientation in San Francisco.
The work. The posting names "Designing evaluations and RL environments that test models in your field systematically." Fellows also team up to build environments that test Claude on judgment-heavy problems, and work alongside Anthropic research teams.
Who it asks for. The posting asks for a PhD in philosophy, cognitive science, economics, policy or law, or equivalent research experience, and says the PhD can be in progress. It also asks for hands-on experience using Claude and Claude Code, comfort in a fast-paced environment, and existing US work authorization for the whole program. Experience designing evaluations, benchmarks, datasets or expert rubrics is listed as preferred.
What stops you. Work authorization. The posting says Anthropic is not currently able to sponsor visas for fellows, and that candidates must have or independently obtain full-time work authorization in the United States.
What to practice. An RL environment with a verifier is a task plus a checker that scores an answer. Our tutorial on building an RL environment with a verifier in Python builds one from scratch. It teaches the mechanics. The posting's emphasis is on judgment-heavy problems in your own field, where deciding what the checker should accept is the hard part.
This is one lab and one posting. Nothing here says you will be selected, and the page can change or close. Check the posting before you plan around any date or number above.
