Ermes Slavoi

[Ermes Slavoi @ESlavoi9987]

@SameBeni @TwtZey Perisa took everything from assyria they copy and steal everything from them . So no not perisa or greek only mesopotamia by long shot then Egypt then the rest

[Ermes Slavoi @ESlavoi9987]

@theo I saw them on my feed all the time

[Ermes Slavoi @ESlavoi9987]

@pranavxmeta @xinhash But this will not consume less ram !? Its just a safari browser no matters how much you do on top of webkit safari just does ot better and more efficient and u will not save any resources, and nobody here can re invent the engine

[Ermes Slavoi @ESlavoi9987]

@dabit3 Building agi level of control of systems, and already achived way better results than codex computer use level of control and how many steps does it take to do them and how accurate the control and the automation are , so i only need the compute for more researches

[Ermes Slavoi @ESlavoi9987]

@coastyai lol grok take the shortcut copy+paste

[Ermes Slavoi @ESlavoi9987]

@Ananth7e The companies that seems to know how to improve are antropic and openai , everyone else trying and failing again and again and this pattern keep going till they just accept to distill these models and try improving on them like kimi and glm 5.3 flash

[Ermes Slavoi @ESlavoi9987]

@Ud4y_G @Bhavani_00007 Non of the grok model is even close to 4.5 opus let alone 5

[Ermes Slavoi @ESlavoi9987]

@JonKurshita @pankajkumar_dev Half the price and almost as good

[Ermes Slavoi @ESlavoi9987]

@thdxr Can you tell me what the model that big pickle is trained on ? And why its so fast and good for simple things ?

[Ermes Slavoi @ESlavoi9987]

@ibocodes @duummbass I have 1 year student sub and i literally never touched gemini only like when every new model drops testing on simple things get disappointed after couple of iteration than make the last one worse and worse and can't improve anything only for the 40 try

[Ermes Slavoi @ESlavoi9987]

@mthsfy @leodev @nahcrof You mean high

[Ermes Slavoi @ESlavoi9987]

@IndependentEco Nothing new , just hate how i see the word generous with codex everywhere when they are very bad specially for rhe 20$ one completely useless

[Ermes Slavoi @ESlavoi9987]

@samirettali I feel its more than 3x for me with opus vs sol

[Ermes Slavoi @ESlavoi9987]

@MatthewBerman @bot @spacexai Building agi level of control of systems, and already achived way better results than codex computer use level of control and how many steps does it take to do them and how accurate the control and the automation , so i only need the compute for more tests

[Ermes Slavoi @ESlavoi9987]

@MatthewBerman @bot @spacexai I am building a better harness than what already existed , need this sub for testing

[Ermes Slavoi @ESlavoi9987]

@MatthewBerman @bot @spacexai I want to use it for Research of my project

[Ermes Slavoi @ESlavoi9987]

@yangzhong01 @gamerPsychopath @ebenezerDN Nothing funny , i am talking serious

[Ermes Slavoi @ESlavoi9987]

@Da7_Tech I remeber testing their old swe 1.5 model and before it and they were genuinely very different from any other models and extremely terrible, so iam happy to find out that they trained on base k3 which on my uses was just fable 5 distilled model both give me the exact same result

[Ermes Slavoi @ESlavoi9987]

@yangzhong01 @gamerPsychopath @ebenezerDN Well i can use opus 5 which is close to astra and for unlimited time on 20$ unlike codex 20$ not even 30m of use