Its probably being poisoned by its own bad input being fed back into it
Might be time to ditch chatgpt and start doing it yourself :p
#is Chat GPT getting worse (dumber, less accurate)
dense steeple
latent jewel
That's a typical thing llms are bad at and it's unreasonable to expect a correct response from them. When an llm generates the next token, it doesn't know how many tokens(yet alone words) it generated so far. And even if it knew, telling which tokens are relevant to the counting would be another challenge of its own.
Use the technology correctly and you wouldn't be disappointed with it getting "dumber".
potent drift
Nah. 4o is legit. Has been building me great foudnational scripts for unity.
plucky chasm
they all suck in their own unique way. i use 5 llm and, at a glance, when 4o blows, claude or sometimes gemini 1.5pro catch the slack