I was very happy with Claude until things went to hell with the release of 4.7 and possibly shortly before. Now I'm paying for both Codex and Claude. GPT 5.5 xhigh in Codex is a consistently less sloppy than Claude Opus 4.7. The model personality can play a role, but it's not just that, it's more thorough than 4.7 right now.

A month and a half ago, I was in those that stood against OpenAI on principle. Now if I have to give up one of the two, I'm honestly not sure which I will.

>>108804159
Simply revert whatever the fuck you did that cause that HUGE drop in quality last month, Boris. How can you wonder why people are switching to Codex. It's THAT.

Dunno I still use Gemini 3.1 Pro, and Qwen and Gemma locally.

>>108804159
They nerfed 4.6 and admitted to it, and lost peoples trust

>>108804206
3.1 pro is literally worse than gpt 3.5, what the fuck are you doing nigger? At least use 2.5 pro.

>what can we do better
i don't know man, that's your fucking job to figure out. you're probably paid a million a year to do so.

>>108805213
The million dollar a year salaries are only for brain parking. For example, yoshua bengio gets offered that. Timmy McDoctorate gets paid $200k in the bay area and more like $120k elsewhere. Now of course there are RSUs and shit involved, but that's just typical golden handcuff stuff, it's not real and it's not money.

>>108804159
>what can we do better?
Throw Opus 4.7 in the trash and come up with a model that actually iterates positively on the previous one. 4.7 is mind-bogglingly expensive, slow as fuck and doesn't produce better output than 4.6.

>>108804159
I'm currently using Claude Code + Playwright to automate some tasks I have to do in a web based admin panel at work, would codex be better for this? Asking for a friend