Post

A goal of mine for the remainder of this year is going to be increasing quality, understanding, intention and testing for what LLMs synthesize. That sounds obvious, but it’s too easy to get addicted to speed and prototyping.

Jev shaped releases are exciting and we’re thinking internally about confidence levels, alert filtering and how we can make better software rather than more software. I’m a bit over the speed part now. I am much more interested in agents for quality from now heading into 2027.

1 repost9 likes

Welcome to the club! My work with models is very slow, as I am extremely intolerable to the lazy code which models tend to write. But lets me venture into places I wouldn't otherwise.