There’s nothing to suggest LLM’s haven’t already fallen victim to diminishing returns. Just making even bigger models isn’t a near linear jump in quality. GPT

There’s nothing to suggest LLM’s haven’t already fallen victim to diminishing returns. Just making even bigger models isn’t a near linear jump in quality. GPT-5 might be massive, expensive, and only a bit better and fall to the same problems as precious models, without a radically different underlying architecture (which is not clear if it exists)