GPT-4 is getting worse and worse every single update

I cancelled my subscription 2 days ago.

About 3-4 months ago, when they introduced custom instruction, GPT was the best thing ever, a great ally with whom I felt confident exploring new API, refactoring code, generating tests, implementation suggestions, etc.

Now it is even worst than GPT3 in terms of hallucinations, stupid+noisy “summary” even when I explicitly ask for code answer only, wont follow A GODDAMN instruction anymore; I made a custom language to rapidly interact with it which worked neatly 3-4 monts ago eg.:

“/tb+5” = “make a pro/cons table with 5 rows”, or

“/fs.ws.x++.cm-” = “(fs) use fsharp, (ws) following scott wlaschin recommendations, (x) explain the code (++) in great detail, add (-) concise (cm) comments”

But worst… it won’t write 1/4 of the implementation code it used to… “implementation placeholder comments” should be HARSHLY punishing for the model, it’s now too complaisant.

I was moving fast, it was fun, rarely needed to reformulate my instructions/questions or, most of the times when I had to do it, the one to blame was me.

Now I waste HALF OF MY TIME (no exageration) reformulating, re-asking for the same damn things, repeating again & again instructions FOR EACH PROMPT (since it wont remember a damn thing) this is totally awful!!

I’m happy with my cancellation because OpenAI doesn’t seem to respect its “Plus” Users, why not rollback the damn thing?!! Why no option to choose the model I was using 3-4 months ago?!!

This is BS… if nothing is done, I hope Gemini crush ChatGPT; until then, I won’t re-subscribe for such a bad product/service.

I encourage everyone here to unsubscribe, this is our only way to get attention

Hang on everybody, this is just a REALLY bad phase…

7 Likes