Skip to content

Anthropic

Anthropic invites philosophers to sharpen Claude’s reasoning

2 min read

Written by an AI model from the linked sources and checked against them by a second model. How we work

Image: pexels.com

Anthropic is taking applications for two new fellowships until October 28, it announced on X on October 15. If you work in science or engineering, the track is STEM Fellows; if your field is philosophy, cognitive science, economics, policy or law, it is Conceptual Reasoning Fellows. Either way it means four months at Anthropic from January 2027, helping Claude reason well about some of the hardest problems in your field.

In Anthropic’s job listings for STEM and Conceptual Reasoning fellows, pay is $100,000 to $167,000 for the four months, the equivalent of $300,000 to $500,000 annualized, plus other benefits. The fellowship is full-time and runs from January 12 to May 8, 2027, and some fellowships may be considered for an extension. Later stages of selection typically include a take-home exercise and conversations with the research team, and offers are planned by November 20. For the STEM track, prior machine learning experience is helpful but not required.

Conceptual Reasoning fellows work on questions with “no single checkable answer.” They design evaluations and RL environments, task setups in which a model learns from rewards, that turn their expertise into “working tests of how well Claude reasons in your field.” Representative projects include a philosopher building hard cases that require normative judgment in ethics and a policy scholar drafting regulatory and cost-benefit analysis tasks with expert rubrics.

On the STEM side, small pods of fellows from related fields meet weekly with Anthropic researchers. They define the hardest open problems in their discipline, build the datasets and simulations to test and train Claude on them, and look for the places where models are “confidently wrong.”

A first cohort of more than 20 STEM Fellows, across mathematics, physics, the life sciences and engineering, ran from June to September 2026, and the problems they built already feed Claude’s training and one of Anthropic’s internal science evaluations; the size of the January cohorts has not been disclosed.

In the STEM listing, Anthropic says it is working toward “an AI scientist, a system capable of the long-horizon reasoning and experimental judgment” needed for challenges such as resolving long-open problems in mathematics and designing medicines for neglected diseases, which “require deep expert judgment that today’s AI still lacks.”

No comments yet. Be the first.

Join the conversation

Sign in to leave a comment. It appears after moderation.