Me too! Finally, these LLMs are showing some appreciation for blunt and concise answers.

Something that used to annoy me about all previous models is that if I asked for a fix to something in a code file (i.e.: fix this method in this class), invariably they would return the entire thing with a bunch of small edits.

GPT 5 is the first model I've used that has consistently done as it is told and returned only the changes.