CasualArthur

[CasualArthur @CasualArthur]

@balog26978 @Pratyusshmd @johnsjohnxva2 @pradeepXkapoor If by "sustainable" you are asking whether you will hit a limit by chatting with it, no, you will not. I heard that it's also a much better communicator than Opus 5, but I personally haven't tested it

[CasualArthur @CasualArthur]

@Pratyusshmd @johnsjohnxva2 @pradeepXkapoor Yes, same for me on pro. I think the difference between it and Opus 5 is quite noticeable. It makes way fewer mistakes and it's a lot more thorough, but that seems to come at the cost of taking much longer per task than Opus 5

[CasualArthur @CasualArthur]

@pminusq @haider1 A counterexample would, otherwise it's just a confirmation of mathematicians' beliefs

[CasualArthur @CasualArthur]

@gurtej__gill_ @rynorhn How is OpenAI falling behind? Astra is the most capable public model and they have developed an internal model, which they claim to be "significantly more capable than GPT-6 Astra" (obviously should be taken with a grain of salt, but if true, that puts them quite ahead of others)

[CasualArthur @CasualArthur]

@oudmeth @JamesSonicemi @synthwavedd Definitely a bot, look at the profile pic

[CasualArthur @CasualArthur]

@jj3maj @OpenAI Hmm I would agree to go 50/50 with you. I bought an Nvidia GPU two years ago, which helped Nvidia invest into OAI, and that money was used to solve this problem. I feel like my contribution to the solution cannot go unnoticed

[CasualArthur @CasualArthur]

@Modernengineer @OpenAI What's insane is the speed at which this improvement was made, considering Astra was released 5 days ago

[CasualArthur @CasualArthur]

@jj3maj @OpenAI Read the article, they decided not to claim the prize money

[CasualArthur @CasualArthur]

@rapquinones @goodhunt N=1, obviously