By skyforbes I'm learning a foreign language and I've been using ChatGPT daily to review a journal entry in my target language. I developed a minimal set of instructions about six months back (maybe more) and things have been working great ever since. It's survived several model upgrades through 4.* and even 5 and 5.1. I've been using 5.1 for a few days now and didn't notice any change.
But today, it's just … broken. Its suggestions are wrong. It's pointing out that I said something correctly. It gives contradictory corrections. It suggests that I change my entry to be exactly the same as what I typed originally. It makes inane stylistic suggestions.
I've spent the last few hours trying to adjust my instructions to make it useful for this purpose again, and it's really super annoying.
Did anyone else notice a recent change that resulted in similar degraded language cognition?
Anyway, I'm off to try LLMs from other vendors.
Very sad.