@zeeg@aboodman@artman I find agents produce poor abstraction that not gonna survive a few iterations, hence the produced "abstraction" is really just for 1-2 use cases
@mitsuhiko This is so true. I feel that the slop being put out without careful thought and hands on being fragile and fall apart quickly. I read the code, too. But that does not stop it from happening.
Intelligence, logic, and common sense. π§
True intelligence is using critical thinking and judgment to reason your way to a consistent answer.
LLMs might give different answers to the same question, or use conflicting reasoning to reach the same result. Thatβs repetition.
@cj_pais Handy is such a good app, using that from the early day a year ago. I was worried how would you afford living building such an amazing app and open-source it!
With the recent GitHub Copilot pricing changes, it's clear that they were suffering from losses because the previous pricing structure was too good to be true. It has also become clear that token pricing efficiency is more of a problem than ever.