[mpalmer]

It's really striking how similar your description is of your progress as a writer to the progression of LLM quality, particularly in terms of long context windows ("the more I wrote, the weirder they became") and training ("I extracted what I needed").

At least one key difference is your ability to self-evaluate. That reflexive ability to realize your own shortcomings over time is why human intelligence is still more of a flywheel than that of machines...for now!

[hoaitx]

Oh, you've pointed out something interesting to me. Indeed, I'm honing my writing skills by writing more and reviewing them more often. It's like an endless loop.