NLConceptual Reasoning Researcher
Trajectory Labs, PBCRemote
Global catastrophic risksAI safetyResearchSoftware engineering
Why this is on the board
Anyone who has AI models help think through hard questions, such as how to keep powerful AI safe and controllable, needs to know where those models sound convincing but reason badly. You build the tests that make this visible: one good test measures every subsequent model again, though safety teams and policymakers decide for themselves what to do with the results.
This board is in beta and this is our judgement. If you think we have it wrong, tell us. Give feedback · How we reason about it
About the role
Design AI safety evaluation tasks with rubrics that measure conceptual reasoning in governance, alignment, macrostrategy, or economics. - Identify hard conceptual questions where frontier models reason poorly; create well-posed problems paired with opinionated rubrics. - Run tasks against frontier models, iterate until scores separate strong reasoning from plausible-sounding weak answers. - Requires original analysis in AI risk, alignment, macrostrategy, economics, or governance with articula…
Want to understand this first?
Start with our free course: about 100 minutes in six short units, at your own pace, with no sign-up. To go further, apply for the EA Fellowship: five weeks of reading and discussion in a small group. Also free, but selective.