@paulg@BuckBundeals Panagram screenshot replies suck and are a waste of space.
Easy way to feel holier than thou while outsourcing your thinking to an LLM too.
@LoganGrafTax This happened to us in the middle of the night, it sounded like someone was breaking in and I got soaked when I ran into the bathroom.
You can buy full metal supply hoses, needless to say everything in my house has them now.
@Stellan628605 With the old brass body faucets, if you open the hot side just a little bit, the stem will expand due to the hot water flowing and close down on the washer again
@emollick Do you have an impression of the best model variant and effort setting to use? With codex recently, I had stabilized on always 5.5 xhigh for any coding task, looking to jumpstart my intuition for best settings to try across the {sol, terra, luna} x effort matrix
@zeeg Even prompt engineering has title inflation. You used to be able to tell the model to think like a staff engineer, now you have to say principal engineer π
@CyclingYazz The colors for watch score overlap with the colors of the mountain/hilly/TT icons, so itβs hard to visually skim and know which attribute youβre looking at. Iβd try to separate these by using distinct colors, or color for watch score and representative icons (clock mountain etc)
@Kraxkrokat@andrewqu Similar feelings going back to Opus 4.8. I have to be more specific about scope and follow through (eg "please update the docs to be accurate again after these changes") where Fable seemed to have very good project level judgement.
@threepointone@javivelasco Same experience. My impression is that scaling up agents lets you build a lot of software quickly, but if you want to be specific about what you build, you still need to be in the loop manually supervising one or two agents.