@briannekimmel Yea, policing for the sake of it is annoying, but there’s a legit annoyance because slop is so distracting. It breaks the flow state of reading comprehension and sows doubt about the integrity of the entire piece. Eliminating the obvious AI tells feels like common courtesy now.
@garrytan is CC's native Plan Mode supposed to engage during the GStack flow? Just happened a couple times and it seemed to siphon off GStack plan updates from CEO/Eng/Design reviews into a separate CC planning doc. Implementing the GStack plan was therefore pretty incomplete.
@benthompson Seems super similar to the overzealous 2FA suggestion from email/messages that shows up next to every text field of every app and you can’t get it to go away.
@conductor_build noticing that the default setting of having Haiku handle PR creation is instantly triggering chat compaction now that Opus' context window is 1M vs Haiku remaining at 200k.
@trq212 Sort of feels rewarding when you see it trigger in the CLI. Makes the session’s work feel more constructive in some way that it’s being proactively written to memory. Would like seeing it more often, or maybe just more visibly.
@bcherny Is it intentional for the main chat thread (in desktop app) to be so greyed out (and therefore hard to read) when the brainstorm skill presents you with the option selector? It’s really hard to get context on which option is right when the rich context in the thread is obscured.