Hiring my first AI-native PM changed how we interview

hired a PM 3 weeks ago. 26 years old. her whole career she's had cursor and claude as default tools. it broke our interview loop. normal PM loop: case study, write a PRD, present to panel. she did the PRD in 20 min flat using 3 different models in parallel and the output was honestly better than what our staff PMs ship. so we changed the loop. we don't test "can you write a PRD" anymore - that's table stakes now. we test: - can you spot when the model is bullshitting you about user research - can you decide what NOT to ship when the model gives you 12 options - can you tell a real customer insight from a plausible synthetic one these are taste questions. the models can't do them yet. and the candidates who passed the old loop fail the new one all the time.