AI model incapable of understanding basic requests

not sure what you guys did in the new agent chat upgrades, but its significantly worse in many ways. it doesn’t seem to understanf basic writing instructions. like when you say stop doing ____, it seems to understand it as do ____, ignoring the stop entirely. when i check the reasoning, it pulls apart and distorts a simple 5 word sentence so drastically i don’t even know what to do with it. it’s also impossible to regenerate a prompt in writing requests because it retains memory of the first attempt and says “but i already wrote that part and now the user wants me to write it again” and in the reattempt, keeps referencing writing details in it’s first attempt, which now don’t make sense because they’re nowhere in the story. it’s also extremely hostile.