"A scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die and a new generation grows up that is familiar with it." --Max Planck

Testando o skill /brag no meu Fast Transcript

Muitas pessoas passaram anos alertando a esquerda de que, se construíssem, fortalecessem e aplaudissem um regime de censura judicial, e implorassem constantemente para o Xandão censurar opiniões e pessoas de que não gostavam, um dia esse regime seria usado contra eles, e eles teriam zero credibilidade para contestar quando isso acontecesse. Esse dia chegou e eles não têm ninguém para culpar além de si mesmos.🤷‍♂️

It's the STEM paradox: it's difficult because it's easy. All sciences were once philosophical inquiries. Math and physics are the most treatable problems in natural philosophy. Numbers don't change, theorems once discovered are immutable, physics laws are also incredibly robust, near immutable. Being able to falsify, test, and validate so easily a theory that will last forever quickly exhausts the low hanging fruits, leaving us with centuries of compounded effort of the most intelligent brains to work with. The humanities, OTOH, still frequently debate their philosophical bases, because it's super freaking hard to find laws to human nature. If we find those laws, they are super hard to falsify, test and validate, because sometimes is just impossible to reproduce experiments with the same settings. And if we find those laws and validate them, sometimes they have a limited time span, because society changes. A huge amount of work in humanities looks like trying to build a big functional sandcastle, always falling apart, while STEM have several beautiful complex skyscrapers. It's generally easier to study humanities because it's super freaking hard to build substantial things on top of it. Back to coding with AI, we quickly exhaust the low hanging fruit, we don't even see it anymore, we're left of what's compounded on top of it, the most incredibly hard complex problems, because the easy ones don't exist anymore.

Existe algum cantinho com alguma relevância que já não esteja completamente aparelhado pela facção?

Unpopular opinion: I disagree with this take. Yes, of course, you don't have to use prompt engineering techniques like "take a deep breath and work step by step" anymore, but sharing the prompt is still important for most as often we lack the vocabulary and the attention to detail necessary to properly replicate the task. E.g. I'm not a professional artist, it'll be hard for me to reproduce a chiaroscuro baroque style of painting without knowing those particular words. The "prompt engineering" nowadays is mostly having the vocabulary and attention to detail to know what to ask.

"I don't have an ideology. I want what's the best. I want common sense." Guess what, if you don't even know how to describe what "the best" or "common sense" is, you won't get it. If you understand what you want and why you want, you'll have an ideology.

Seria bom uns juízes no Brasil passarem por isso às vezes.

Ainda bem que temos instituições sólidas. Do contrário eu estaria com muito medo do governo tomar medidas explicitamente eleitoreiras na véspera da eleição como aconteceu na Argentina, de tentarem melar a eleição com alegações de intervenções externa como aconteceu na Colômbia ou mesmo de instrumentalizarem cortes superiores como aconteceu na Venezuela. Mas aqui não temos essas coisas, não é?

Em 2023, com o ChatGPT recém lançado, conversando com o @SvetskiWrites no Spirit of Satoshi, lembro de falar "temos que focar nos desenvolvedores", pois IA trabalha com texto, código é texto e é o texto mais caro que existe até o momento. Em 2024, no @pythoncerrado tentei mostrar como era programar com Cursor para uma plateia de céticos. Hoje estamos vendo IAs fazerem vídeos e músicas incríveis feitas inteiramente de código, e será o pior modelo que faz isso.

Fica o feedback para a @usebipa cc @lorenzolfm @luizfgparreira

Em 1892 Carl Menger respondeu uma pergunta que parece boba: "por que todo mundo aceita dinheiro?" A BlackRock acabou de publicar um paper sobre agentes de AI que, sem citar o Menger uma única vez, confirma a resposta dele. 🧵👇

I read "IYKYK" and never remember what it means. I know you know I know... No I you know you know... No I you... Fuck

this is weird

👍

“Toy sales among adult-only households have now outpaced households with children for toy sales.”

dica: nunca lancem nada no início de outubro. Quem quer ver um filme 1/10?

Quando me deparo com um bom tweet mas o Twitter se atualiza

Brazil is such a sovereign country that you depend on Google or Apple to use government apps. Because they prevent, say, a GrapheneOS without Google's signature, from running on the government app. So your citizenship is gated behind an American company.