If it had been only a year they’ve been saying that, I might almost buy the logic. But it has been 4 years since Chat-GPT came out and they’ve been making the same promises the entire time. Something of a tangent… thinking back to 2022-2023 my expectations for what the LLM companies would actually achieve by now were actually too high? Like I would have thought they would have figured out a better architecture for gluing functionality and symbolic logic to LLMs than just calling tools with the output text stream and feeding the tool outputs into the context window.