Would it be useful to collect the contexts, where various LLMs think the same? — LessWrong