One result is a flare, not a sunrise. Surprising claims (delightful or uncomfortable) earn replication. Make round two stricter: preregister, tighten measures, reduce researcher degrees of freedom, vary the setting or sample. If it’s real, it travels. If it’s fragile, you just saved months of theater. Build replication into the roadmap like QA: default, not punishment. And when a headline result lands on your desk, trade awe for discipline. Invite independent repeats. Evidence that keeps its shape across time, teams, and contexts is evidence you can build on.
