I have always believed that you don't need a GPT-6 quality base model to achieve human-level reasoning performance, and that reinforcement learning was the missing ingredient on the path to AGI. Today, we have the proof -- o1. x.com/OpenAI/status/…

Think Hard.

think slow.

O yeah there's no waitlist

o1 is not GPT. your prompts will not "just work". you'll have to experiment to find where it shines. you'll likely find ways to use it that even we don't know about. go play!

y'all just got bitter lesson'd

ty for all the feedback so far! few notes: 1. yes we went out to 100% of plus users yesterday. this is my preferred way of shipping, even if there’s some risk demand is overwhelming 2. doing what we can about rate limits! for now we reset them because there’s so much enthusiasm

o1 prompting is alien to me. Its thinking, gloriously effective at times, is also dreamlike and unamenable to advice. Just say what you want and pray. Any notes on “how” will be followed with the diligence of a brilliant intern on ketamine.

config is an AGI-complete problem

selling shovels doesn't work when digging gold also makes gold dig itself

everyone glorifies silence and lack of distractions as the only way to focus and get things done. skill issue. you should be cranking code for 3 different projects while on a large coffee listening to industrial noise pop at 260 bpm

one day i'm going to fix tokenization just so you guys stop getting nerd sniped by it when prompt engineering

the long tail is long

an incredible amount of care and craft went into shipping this amazing experience

the knowledge industry is such a weird misnomer. it was never supposed to be about what you know, but about how you think

everyone wants to join a rocket ship until they remember they’re bumpy and on fire. buckle up y’all, only way to space.