Are you polite to your LLM? I catch myself writing "please" all the time and wonder if it's a waste of time. But then again, a true gentleman behaves even if nobody is looking, right? So let's not forget our manners.
That said, on the other extreme, I want to yell at my LLMs for not coming up with the one line fix in 40 round trips and hours of wasted time but as they have no memory, it's pointless to tell them where they are wrong.
Surprised that reasoning models didn’t exist earlier yet. LLMs are glorified data processing pipelines, and you can only fit so much processing in a set of layers (I would be willing to bet most of them internally converge to the following architecture: decode => process => lookup-like mapping => process => encode)
Reasoning (with scratch space) compared to without is like combinational logic compared to sequential.
I went through a drive through with an AI order taker and I didn't say thank you after completing my order. I'm going to hell.
