GPT-5.2 does not follow instructions and ignores my prompts

People in the comments are all talking about a degradation in performance, but I used GPT-5.2-Thinking for some searches yesterday and it followed my instructions perfectly. There's no way to know for sure if the model is experiencing issues, that's the problem with this non-deterministic tech.

When they first upgraded, I saw that problem over a span of a few days. I gave lots of feedback because the problems were easily reproducible.

For example, ask for an image, with specific set of details, in 16:9 format, would present 3 images that completely ignored most of the details, and none would be in 16:9. I would have to downvote and "scold" it often to get the single pics with the content I was asking for.

This problem went away after maybe a week. I haven't really seen anything too annoying since. However, even though it claims it can do a full series in one fell swoop, I don't trust that enough to bother trying. One-by-one it is...

I also specify .jpg, but it switches format randomly sometimes. I know .webp is "better" but I'm tired of explaining what webp is to people, and dealing with file managers that don't show the format in tiles. So there are a subset of my target that only get slobberproof content from me.

Tools for fools.