Gpt and AI Discussions

Message

#136 Post by **ViperDiscord** » Mon Mar 30, 2026 11:20 pm

One thing on codex I've noticed.. and I could be wrong.. but sometimes it just seems to use a lower model without telling you. Its very obvious when it happens because the replies it gives and the code it creates are pure garbage.

#137 Post by **ViperDiscord** » Mon Mar 30, 2026 11:22 pm

I've been using codex high. It does great for small stuff, but kinda goes really crazy with large tasks. I wanted to add some new api hooks for other devs and it created like 20 classes lol. Claude did it in 3

#138 Post by **ViperDiscord** » Mon Mar 30, 2026 11:23 pm

They ALL suffer from code bloat. If I ask it to update a single function it wants to create 2 helper functions to do it.

#139 Post by **ViperDiscord** » Mon Mar 30, 2026 11:23 pm

Yea, it's really annoying

#140 Post by **ViperDiscord** » Mon Mar 30, 2026 11:24 pm

Lol its like they get paid by the function.. it wants to spam them like crazy.

#141 Post by **ViperDiscord** » Mon Mar 30, 2026 11:25 pm

Days when devs were judged on how much work the got done by the lines of code they added

#142 Post by **ViperDiscord** » Mon Mar 30, 2026 11:25 pm

I've started to tell it give a minimal diff to do xyz. Then it starts to limit extra function creation.

#143 Post by **ViperDiscord** » Mon Mar 30, 2026 11:26 pm

I always have it do a cleanup pass to reduce code and that usually helps.

#144 Post by **ViperDiscord** » Mon Mar 30, 2026 11:28 pm

For me its speed. I'm a big speed optimization guy. I cant stand code that is not tuned to be highly optimized. And it gets very frustrating having to go and clean up various things. Its crazy but its one of my OCD things I just cannot help. Started in colledge with a parallel processing class, you got extra points for who's ever code ran the fastest.

#145 Post by **ViperDiscord** » Mon Mar 30, 2026 11:29 pm

I've probably said this before but I will spend hours tuning a single function to get just a few ms shaved off of it.

#146 Post by **ViperDiscord** » Mon Mar 30, 2026 11:30 pm

When you are dealing with a GOM that has 50k objects in it and you are are operating OOP, you cant just say hey lets loop through the entire GOM to grab this item, or whatever.

#147 Post by **ViperDiscord** » Mon Mar 30, 2026 11:31 pm

Oh that's, neat.
The stuff I'm working on is rendering so it has to be fast.

#148 Post by **ViperDiscord** » Mon Mar 30, 2026 11:31 pm

Yea 50k is a lot

#149 Post by **ViperDiscord** » Mon Mar 30, 2026 11:33 pm

So you understand dealing with rendering. Every ms you can shave off is that much more fps.

#150 Post by **ViperDiscord** » Mon Mar 30, 2026 11:36 pm

Gpt and AI Discussions

PitViper Wrote:

Mysticdrew Wrote:

PitViper Wrote:

Mysticdrew Wrote:

PitViper Wrote:

Mysticdrew Wrote:

PitViper Wrote:

Mysticdrew Wrote:

PitViper Wrote:

PitViper Wrote:

PitViper Wrote:

Mysticdrew Wrote:

Mysticdrew Wrote:

PitViper Wrote:

Mysticdrew Wrote: