Which is yet another reason why the current bubble is so insane… Big tech is making several hundred billion dollar bets that more text means more smart, based on utterly irrational optimism despite evidence to the contrary

We have never had the opportunity to create a technology like this before, and it's quite possible that after we've created this technology, the opportunity will never exist again 😅

I think it's already peaked… the funny thing about LLMs is they sow the seeds of their own destruction 😅 they require an enormous corpus of text written by humans! and then once they are created, they start destroying the very corpus of text that "created" them… It's unprecedented in technology

So far, I have found those problems better solved by rule based systems like regex, refactoring out boilerplate, or more powerful editing techniques like vim, multi-cursor, etc. I find that LLMs sit in an uncanny valley of ineptitude where they aren't good at creative tasks, nor structural tasks

This is why I don't use it — I enjoy writing code, and unless you process information, you are not actually learning. Recognition provides the illusion of learning while recalling and processing provides deep learning. You've got to write code, understand, build pattern muscle memory, then level up

One habit I'm really proud of my team for adopting is asking "goofy questions" — I try to tell people I'm not actually smart, I'm just stupid _faster_. So I like to draw attention to my benign ignorances and resultant curiosity with "This is a goofy question, but..."

Now everyone does it 😁😁😁