I have had early access to OpenAI's GPT-5 for the last two weeks and it is the smartest model available as of now. Just as one example, here it created an anamorphic text illusion in minecraft that spells "SPECIAL" from one angle and "GPTFIVE" from another

I was about 2 months early. In fact, this was before the Dwarkesh post.

on one hand, there's nothing mindblowing or overly splashy. but on the whole it seems like an incredibly useful release. i expect this will be my daily driver for both chat and coding, and it'll be a bit better across the board, less frustration and hallucinations. i'll take it.

Did it deliver on the 1 yr+ hype? Seems no. Was the livestream well-executed? Far from it. Are you guys being complete haters? Yes. We got a very useful model (system) for very very cheap today.

“Is that a PhD in your pocket or are you just happy to see me”

Get ready for the next OpenAI launch to be a sanitized corporate regurgitation slop fest bcz of today's reaction. They were raw, edgy and a little cool and uncaring, but you folks will push them to decide what the CEO should post before a launch via committee and vet the deck

A bit sad how the GPT-5 launch is going so far, especially after the long wait and high expectations - The automatic switching between models (the router) seems partly broken/unreliable - It's unclear exactly which model you're actually interacting with (standard or mini,

I only stay on this shitty app bcz there is no other place for me to interact with the kind of people which are here. If Zuck had done the threads launch right by not letting the IG low-iq crowd on the app and by not linking profiles with IG (not anon), we all wd have left.

Bros, GPT 5 is a good model and a decent release (minus the launch event ig). Don't hate on it.

weird seeing people say they miss 4o since I hadn't used it in months. o3 and 4.5 guy altogether. miss o3 a bit but until now GPT 5 Thinking has been reliable.

Lately I had been started to feel that I was losing my ability to think bcz of reliance on LLMs. Now I'm trying reading books often and writing and taking notes religiously and oh boy does it really, really help in restoring my thinking faculties.

When we got o3, I had to tweak my system prompt in Chat multiple times until I got good results. And this was because it was much better at instruction following that previous models. GPT 5 Thinking is something else. Seems to do everything you ask. Incredible!

Remember when we used to pay $20 a month for GPT-4???

As it stands today, small models can be more valuable to businesses than larger models.

This makes me want to kill myself.

The ability of a model to produce great performance under any scaffolding should be measured. Smart people are adaptable and smart models should be too.

I talk to many folks who constantly think about the mistakes they've made in the past and are regretful. But why not derive initiative from this exercise? You should've learnt to code? Do it now! You think your ex would've respected you if you didn't shout? Make the change now!

Spit-bench: where LLMs write bars and compete with great verses from human rappers.

GPT-5 says 'I don't know'. Love this, thank you.

It's hilarious how no big lab provides a chat UI with Branching of conversations visualised. In fact, only Google provides an in-built branching option. I'm tired of tracking my branches, guys!