Yep. In my experience, LLMs easily go in circles with even simple assembly, creating fixes that result in 2x slower code and then fixing those with even slower code. They are pretty great as a reference or finding needle in the haystack bugs though!
Sure, they might make all sorts of dumb hypotheses at first, but as long as the results of those stay in their context, they do seem to eventually “run out of ways to be stupid.” (Which is to say, LLMs seem to experience in-context learning even via self-directed trial-and-error, if given a sufficiently-large number of iterations and no way to cheat.)
It also is an excellent radare2/rizin/ghidra driver as well. Maybe it's because its ~cyber~ capabilities (pretty much linked with assembly-level knowledge) aren't guardrailed off?