I have a hard time believing whatever prompts get Claude to reason can stay relevant secret sauce for long anyways. It’s not hard to A/B test something that gets you close enough, and it’s not Ike anthropic has uncovered the global optima of reasoning prompts.