> Imagine […] what you could do, ask, plan, or learn, if you had the full undivided attention GPT or Claude. not a commercial, guard railed, fine tunes, beat into submission version that is splintered into hundreds of millions of iterations to chat with every one.

That is not how it works. It is not “splitered”, there is no divided attention.

There is nothing to indicate that LLMs are improving "exponentially" at this point.

The last few iterations show a logarithmic curve at best tbh. If we are to see a major advance, it'll be something like the implementation and infrastructure for byte-level transformers.