@thsottiaux Opus 5 é aquele funcionário que vc pede pra ele cortar a grama e ele volta te falando que ele limpou a piscina e que seria interessante vc podar as folhas da arvores
Parece que tem tdah
To tentando usar ele ao invés do 5.6 Sol que acabou os tokens da semana mas tá um desastre
@PallottaPedro Foi só pros mininos do Lula comprar uma cervejinha na Espanha po!
Só quem nunca levou uma ré violenta dessa por marginais que continua votando nos esquerdosos, seja no brasil ou fora!
Só o método Bukele pra solucionar essa bandidagem solta que os esquerdosos defendem.
@ZixuanLi_ Congrats for the awesome model !!!
Great you are keeping the size of the model and not inflating it.
The only thing left is Vision to make it even greater!
@sudoingX You are so Wrong!
Just remember Apple accusing Openai of stealing business secrets.
Remember that qwen 3.6 27b has today the strenght of a frontier model of 365 days old. Imagine in 365 days having a fable 5 like running locally.
Also that memory prices will eventualy fall.
I solved 6 open Erdős problems in 5 days, using @OpenAI GPT-5.6 Sol.
I have a math background, but the Codex workflow I used does not require deep mathematical knowledge.
Here’s exactly how I approached it, including my prompts 🧵
@elonmusk When 100 Quadrillion model? You really need to be more focused on the mission!
I really need antigravity, wormhole travels and smart robots making me an Enterprise type spaceship landing on my own halo world.
Thanks for your attention.
@MatthewBerman Americans are not used to communism and socialism bureaucracy limiting free market. So I explain how it works to you
The companies will simple leave america or be built somewhere else. It is what always happened, but generally they go to America and now they will leave America
@osanseviero Also specialist models.
Because if we want one model to be specialist in all knowledge, it's size get's astronomically large, the way frontier models are getting
A coding model do not need to know deeply biology, law and finance
Specialization aways wins over generalization
@osanseviero A 300 I.Q / 10B params model would be great
Any model with access to the internet have all the world's information, doesn't matter if is a 1Bi/10Tri model
The diferenciation is how intelligent it is and how dense all that intelligence can be compressed inside the model's size