Mostlime

17 years old
im a cool programmer guy
(+ I also like art, music, and writing :)) )

[Mostlime @Mostlime12195]

@XVPbhwyyKr61371 I just hope that Sonnet 5.5 and Opus 5.5 are good models. With Anthropic it always seems like the .0 models (like Sonnet 4 or 5) are bad while the .5 models (like Sonnet 3.5 and Sonnet 4.5) are great models.

[Mostlime @Mostlime12195]

@shookscooks @larplegends @himanshustwts for reals man

[Mostlime @Mostlime12195]

@larplegends @shookscooks @himanshustwts I've tested 1.2B models better than this crap.

[Mostlime @Mostlime12195]

@XVPbhwyyKr61371 @FirstSquawk Yeah. So much has changed in 3 years. I don't like Dario but people are spreading so much misinformation. The statement that was shared about him saying that open-source AI was going down a dangerous path was from 2023. He probably has a different mindset now, hopefully.

[Mostlime @Mostlime12195]

@XVPbhwyyKr61371 @FirstSquawk Do not heed these clickbait headlines. Dario said that statement 3 years ago before Llama 3 was even a thing. I don't doubt that Dario still "dislikes" open-source but all these accounts are pretending that Dario said that recently. Dario said nothing recently.

[Mostlime @Mostlime12195]

@XVPbhwyyKr61371 @HarshithLucky3 Gemini is incredibly good at vision and it has lots of knowledge, even if it isn't "smart" or very good at instruction following. because of that, even though I hate gemini, I always use it when I need to ask a quick question or need to understand something or for fashion advice

[Mostlime @Mostlime12195]

@XVPbhwyyKr61371 Yeah. I personally haven't used Fable much, but the few interactions I've had with it made me feel like it was AGI for once, or at least close to it. I'm really hoping ChatGPT's new models give me the same feeling!!

[Mostlime @Mostlime12195]

@XVPbhwyyKr61371 I missed fable so much, honestly at this point I'm happy we're getting it back even if I can't use it for coding. It was such a good model.

[Mostlime @Mostlime12195]

@Kyler_Lorin @scaling01 I don't think that's how it works. CoT efficiency comes mostly from post-training, not pre-training.

[Mostlime @Mostlime12195]

@scaling01 they said be there or be square

[Mostlime @Mostlime12195]

@silentvoicebot Don't fall, little boy.

[Mostlime @Mostlime12195]

@momoedeai @NGEScreenshot I don't know what you're talking about but https://t.co/K34x7UXSvn works fine and I came here to say I love you for making it.

[Mostlime @Mostlime12195]

@Sakura_dacc @scaling01 mf talking bout sum "GPT-4o" in the big '26 😂😂😂 go away clanka

[Mostlime @Mostlime12195]

@Mercuriusdream @tugot17 @arthurmensch Indeed, I do too. liquid time-constant networks rule, as do input-dependent convolutions.

[Mostlime @Mostlime12195]

@ClmentHardy3 @Rasmic the original post is obviously sarcastic but either way many people don't agree with your definition of AGI, and you don't get to act like this definition is the objective or best definition of AGI.

[Mostlime @Mostlime12195]

@ClementDelangue @JakeKAllDay @ArtificialAnlys not even remotely comparable

[Mostlime @Mostlime12195]

@TheLouieCo @Kimi_Moonshot hahaha better than fable 5? not even remotely possible

[Mostlime @Mostlime12195]

@jeet_exist @HarshithLucky3 fable 5 at low reasoning is more token efficient and cheaper than gpt 5.5 xhigh while also being stronger

[Mostlime @Mostlime12195]

@CG170995 @guccheeto @rushicrypto people like us as hard to come by, cheers.

[Mostlime @Mostlime12195]

@hurdur4o @ns123abc @Cazador1231 and yet fable is still stronger than it.