model, but I think it's mostly just the fact that I like what sol/astra do by default when things are underspecified (and I have probably been trained to under-specify things that I like how codex handles). opus its a lot of me coming back 20 minutes later and finding a bunch of choices I hate
and I feel like codex does a better job of inferring from the existing choices in the codebase what I would want it to do when adding something. so I usually make a bunch of choices up-front and then let it rip with vague additions after. this doesn't seem to work as well with opus
huh, I feel like opus does a very good job of this, but a friend of mine has the exact same complaint. i wonder if it’s a difference in prompting or a difference in what decisions we want