Post

i'm kinda surprised I didn't find anything reliable, @⁠code-yeongyu/comment-checker from oh-my-opencode is nice, but it warns instead of modifying files i don't wanna tell the AI to remove the ad-libs, just delete them

even the newest models perform so much HYPEMANNERY I ask to fix a bug, the agent leaves a "fixed a bug here, yay me" comment. there's just zero respect for the fourth wall, the conversation about the process leaks to the medium. and then we're all reading that REPEATEDLY.

You might notice a similar thing when you ask AI for some specific design style, let's say brutalism, and then it also makes the text on the page *about brutalism* just so chatty, can't stop itself