Them

LLMs do not change

You

Not with agents and chain of thought. Agents can run for hours continuously

Those are not the same concepts. The model of LLM remain unchanged. Agents running for hours just modify the prompt they input into an LLM model.

I do not care about the rest of discussion, just swoop in to clarify that distinction.