I think I’m ready for Codex
Claude is all over the place right now. On very simple tasks, like move some text on a page, it consistently invents new things to screw with.
Is Codex just hype? Or worth the switch?
@TheAhmadOsman Yah, the future isn’t in a “do everything” agent. Theres a few really solid, Manus is great. But having a bazillion tools clogging context just isn’t the play for any serious operation
@mattstaton@gregpr07 Interesting. I’d built on browser use before, was a bit clunky. But if I can adapt this also for anti-detection, might be cleaner than my browser
Best thing I’ve done this week is use Claude code to rebuild a version of OpenClaw that actually moves the needle.
The idea of a self building AI is cool, but anything beyond basic tasks or research and it’s more frustrating than helpful , or at minimum wildly inefficient
@0xOzyan I just think the wide flexibility sounds great , but the reality is it’ll have repetitive actions. Which get so wildly expensive for it to do the same repeated thing, but start from “I can control a whole computer”, then work its way down to sending slack message. It’s fun tho!
OpenClaw is cool and all
But it’s soooo inefficient and unreliable
It did however give me a decent framework to rebuild something that’s actually efficient and slays
But good on you, spending $37 in Opus tokens just so it’ll send an email