Ann comments on Testing for parallel reasoning in LLMs