Consider the recent advancements in reasoning models, I’d say your method is a bit inefficient ;) It’s equivalent to the LLMs reasoning in the output and not i

Consider the recent advancements in reasoning models, I’d say your method is a bit inefficient ;) It’s equivalent to the LLMs reasoning in the output and not in the latent space before the final output, which gave the rise to the reasoning models we see today. So speaking out loud might not be the best reasoning method ;)