As I suspected, the 90% score on simple bench is a lie. Anyway, I kinda believe we're seeing the beginning of the end of OpenAI, what Sama didn't do that caused top talents to leave,     See More GPT 5 announcement is an embarrassment, even if the model is good, and might even cause dips in AI evaluations in general. OpenAI is bleeding money and the beef with Microsoft doesn't help them. GPT 5 needed to be ground breaking, it isn't. Tbh, deepmind seem to impress me a lot more than the competition even if we don't consider Gemini.    See Less

I'm crying, so happy and FREE Palestine.     See Less

GPT-5 is the best Coding model, period. I am building a desktop app, and currently working on a embedded pdf viewer inside it. No other model could bring it into a barely useable state. GPT5     See More like 2 Prompts to completely fix it. Also the code quality is much better, and none of these annoying git-commit message like Comments.
And finally an openai model with good frontend taste / skill    See Less

they really just be pumping out slop to increase stock value     See Less

👍🏻👍🏻     See Less

Brilliant first impressions, we loose sense of progress due to the competition and constant update of models, but I think the jump from Gpt4 (original release) to this is insane.. It's j     See More e got so much in between that the jump is not as crazy.. Normal people is gona be blown by it    See Less

You need to use gpt5 high. It's vastly superior to what is available in the app.     See Less

Sam Hypeman strikes again     See Less

We were told GPT-5 would be huge, close to AGI, etc etc. Instead we get some of the stuff that’s been in Google’s AI Studio for a year along with another demonstration by influencers tha     See More reate Tetris and Snake again. It still couldn’t create a decent 12-week training schedule for a half marathon, it couldn’t even add some room for notes despite only asking for that, it was unable to generate an interactive website with a csv upload, it couldn’t even analyze a Strava export without retrying and failing 5+ times.    See Less

Looking at Claude Code, I wonder how much the models vs harness / scaffolding is what needs to improve. Given the right context and work step, for coding related tasks...     See Less