Post

This matches my recent experiences. I can ask for things, and the output works, but it's all wrong in the details. Not in a "doesn't fit my preferences" way, but in a "makes the system vastly more complicated and leaks that out to the user" way.

rain 🌦️@sunshowers.io

I gave Opus 5.5, considered the best model these days, a medium-sized task. Even after a full adversarial review and simplify loop to convergence, it produced code that worked but was garbage, with limited use of newtypes and representable invalid states everywhere

Jeremy Morrell

@jeremymorrell.dev

Maybe the system I've been playing with is too new and doesn't have the bones in the right place yet. I have fewer problems working in mature systems. I'm trying to add in guidance and static analysis and loops and adversarial review and that helps, but... I still get a bunch of bad first drafts

13/10/26, 4:27 pm0 reposts1 like

Jeremy Morrell

@jeremymorrell.dev

Also very open to "you're holding it wrong" type suggestions. It's frustratingly close.