Much of Anthropic's work is empirical research and infrastructure where the answer isn't known in advance: you're running experiments whose outcome you can't predict, building systems against requirements that shift as understanding improves, and making claims from noisy data. The temptation in that environment is to treat ambiguity as license for sloppiness - to run the quick experiment and over-read it, to skip the rigor because 'it's just research,' to ship the thing that looks like it works without the discipline to know whether it does. The value the company selects for is the opposite: bringing high craft and scientific rigor precisely because the ground truth is unknown - designing experiments that can actually distinguish hypotheses, being honest about what noisy results do and don't support, building carefully in the fog, and structuring ambiguous problems rather than being paralyzed or careless in them. In behavioral rounds, interviewers look for how you operate when the problem is genuinely unclear. The strong signal is a story where you brought rigor to ambiguity - designed a clean experiment amid noise, distinguished a real result from an artifact, structured a fuzzy problem into something tractable, or resisted the pull to over-claim from thin evidence - and made real progress without pretending to more certainty than you had. The anti-signals are using ambiguity to excuse cutting corners, over-claiming from noisy or underpowered results, analysis paralysis in the face of the unknown, and craft that only shows up when requirements are crisp.