Continuing the topic of what's better: GPT or Claude
This spring, we participated in a cybersecurity competition, and my whole team used Codex
It was GPT-5.4 because 5.5 kept flagging chats as a cybersecurity risk
We finished 7th (among 300+ teams), or 5th after disqualifications
And all of our tasks were solved through Codex
At the beginning, I tried to using Claude, but it barely solved the easiest task and couldn't try different approached autonomously
GPT has always been smarter, but Claude used to win because of its attention to detail. It's gone now
Well, bye bye Claude
I've been using it since February, but over the last month it's started acting like it's still from February
I can't afford to spend the whole day fixing every dumb mistake it makes on its own
Sorry, Sam, I'm back
Is Astra-6 cooking, by the way?