BWWWOY 🌐 1%

A mulher que construiu o ChatGPT saiu da OpenAI, ficou em silêncio por um ano, e o que ela acabou de lançar pode mudar pra sempre como você usa IA no dia a dia. Mira Murati não fundou mais um chatbot. Ela foi atrás do problema que nenhum lab quis resolver: toda IA que existe hoje funciona por turnos. Você digita, espera. O modelo responde, espera. É tentar resolver uma crise por e-mail quando você poderia estar na mesma sala que a pessoa. O que a Thinking Machines lançou hoje acaba com isso. O modelo ouve, vê, fala, pensa e age ao mesmo tempo. Não é um pipeline costurado de componentes. É o modelo em si que foi treinado do zero pra funcionar assim. → Latência de 0,40s por turno. O padrão da indústria é 1 a 2 segundos. → Micro-turnos de 200ms intercalando input e output sem parar → Faz busca, usa ferramentas e gera interface enquanto conversa com você → Percebe quando você hesita e intervém antes de você pedir → Tradução simultânea em tempo real com as duas partes falando A equipe: Mira Murati como CEO (ex-CTO da OpenAI), Soumith Chintala como CTO (criador do PyTorch), e contratações recentes da Meta em percepção multimodal. O ponto técnico que vale gravar: eles citam o "bitter lesson" do Rich Sutton. Interatividade construída por componente externo sempre vai perder pra interatividade nativa ao modelo. Escalar o modelo o torna mais inteligente e mais colaborativo ao mesmo tempo. 822 mil visualizações em 4 horas. a16z comentando. Brasil dormindo. Toda IA que você usa hoje vai parecer e-mail dentro de dois anos. E quem largou na frente dessa corrida não foi OpenAI, Google nem Anthropic. Foi a empresa da mulher que eles deixaram sair.

Damn，Karpathy这条帖子直接把我过去半年的AI工作流全推翻了🤯 大家都在死等更强的模型， 死等更大的上下文窗口， 但Karpathy说，你们全搞错方向了， 现在AI最大的瓶颈，根本不是模型不够聪明， 是我们还在用文本这种最低带宽的方式，跟它沟通。 他推荐了一个所有人今天就能用的trick， 在任何query的最后加一句： "structure your response as HTML" 然后让Claude直接帮你打开， 出来的效果好到离谱， 不仅仅是多了点颜色和排版， 更像是你终于给AI打开了大脑里那片10车道的视觉超级高速公路， 同样的内容，HTML的阅读效率和理解深度，是Markdown的10倍以上， 这简直就是人机交互的真正下一代范式，因为人类的输入和输出偏好，天生就是完全不对称的， 输入最自然的是音频，说话比打字快4倍，思考也更连贯， 输出最擅长的是视觉，我们大脑1/3的皮层，全用来处理视觉信息， 而我们现在，却在用文本这种单车道的土路，双向跑所有的流量， Karpathy画了一条清晰的演进路线： 原始文本 → Markdown → HTML → 交互式神经视频， 我们现在正站在Markdown到HTML的转折点上， 最令人兴奋的是，很多人说HTML费token，生成慢， 但你算一笔账就懂了， 多花2倍的token，换你10倍的阅读速度和理解深度， 这是全世界最划算的交易了吧哈哈， 可惜我们早就被省token的思维绑架了，却忘了人类的时间才是真正的稀缺资源， 还有一个更扎心的认知， Markdown是给AI看的格式， HTML是给人用的格式， AI代理之间沟通，用Markdown甚至JSON都没问题， 但所有最终要给人类消费的东西，都应该切成HTML， 这才是最优的分工， 现在我已经把所有prompt的结尾，都加上了那行字， 做对比用并排表格，做分析用彩色标注，做原型用交互式滑块， AI不再是给我甩一大段干巴巴的文字让我啃， 它直接给我造了一个可交互的视觉思考空间， Karpathy说，人机的心智融合才刚刚开始， 我们根本不用等Neuralink那种脑机接口， 先把HTML用起来，就是当下能摘到的最大最甜的低垂果实🍒 #AI #Karpathy

8 лет назад я решил написать учебник по языку программирования Python для полных новичков с нуля. Спустя пару месяцев работы над текстом, я понял, что текста будет не достаточно, а еще надо разобраться как работает мозг при изучении новых вещей. Это запустило невероятно большую цепь событий. Благодаря @kukutz меня пригласили в Яндекс.Практикум и я поучаствовал в создании курса Бэкенд разработки на Python'е который на сегодняшний день прошло больше 10 тысяч студентов. Но мне и этого было мало. Я хотел сделать курс лучше чем команда профессионалов в самом Яндексе. И больше 4х лет я занимался созданием своего собственного учебного курса. Изучал когнитивистику, методолгию, много новых технологий необходимых для создания проекта. Создал свою собственную платформу, строил маркетинг, а когда появились студенты то обучал и делал практику. Прошло время, я изменился и изменился мир. Я оказался не самым талантливым продажником, мне всегда было интереснее создавать и в этом мое благословление и проклятие. Я решил, что лучшее, что я могу сделать — это открыть и отдать учебный курс в открытый мир и посмотреть что произойдет. Поэтому я собрал все материалы создал новый сайт для контента курса, а видео перенес на Youtube. Когда мы анализировали рынок, то оказалось, что мой курс по объему относится к нише учебных проектов стоимостью от 2700 до 4500. Мы анализировали предложения от всех крупных школ представленных на рынке. Да у конкурентов в пакет входит предложение проверки кода, поддержки, мотивации и унижения учеников. Эти обязательства я на себя не беру. Но в остальном я выкладываю свой курс в открытом виде и бесплатно. Вы можете зайти на сайт pylot.me и начать двигаться ступень за ступенью. Когда-то я вышел один в поле боя с горячим желанием поменять этот рынок. Но компании с большими бюджетами меня победили. Пусть так, но сейчас я открываю всем свой проект. Вам больше не надо брать кредит на обучение, не надо занимать у родителей и друзей. Теперь вы только наедине с собой и своей дисциплиной. Если вы пройдете этот путь, то сможете выучить прекрасный язык Python и научиться всей небходимой базе. Что я хочу взамен? Мне кажется мир станет немного лучше если у вас будет такая возможность. Я не знаю куда заведет меня этот путь. Подозреваю, что может быть это будет просто классное путешествие. Надеюсь мы все вместе станем лучше в процессе. Пожалуйста поставьте лайк и если можете, то сделайте репост.

@s8box is bandana hair limited or not?

April Community Cosmetics! 🪝📿 Sale ends May 1st 2026, after which it'll only be available on the Community Marketplace. Sales proceeds go towards the cosmetic item's creators, the playfund and engine development. Learn more here: sbox.game/metrics/finance

Every time you see a startup with “.ai” in its domain… someone is getting paid. Not the founders. Not the VCs. A tiny Caribbean island is. → Anguilla Back in 1995, countries got their domain codes: > India got .in > The UK got .uk > The US got .us > Anguilla got .ai At the time, it meant nothing. No tech scene. No startups. Just tourism. Then ChatGPT happened. And suddenly: > .ai started to mean credibility > .ai started to mean cutting-edge tech > .ai started to mean something investors take seriously So every startup rushed to buy it. The result? > Domains grew from ~60K to 1M+ > ~2,000 new registrations per day > ~$130 per domain (2 years) And here’s the crazy part: > Nearly HALF of Anguilla’s national budget now comes from .ai domains That money is funding: > Lower taxes > Free healthcare for kids > A brand new international airport No strategy. No master plan. Just… two letters assigned by chance. Sometimes, the biggest leverage in the world is pure luck.

You're prompting Claude entirely wrong. Because you shouldn't be prompting at all. Do this: 1. Folder ☑ A folder on your computer, "Claude Cowork": Every file inside replaces a prompt you'd type. ↳ No more "here's the context again" in every chat. 2. ABOUT ME/ ☑ about-me .md & anti-ai-writing-style .md file. Who you are. How you write. What you'd never say. 80% of these files should be what you REJECT. ↳ Claude reads them before every task. 3. PROJECTS/ ☑ One subfolder/ project. Inside: brief .md, references/, drafts/ Prompt: "Read everything in PROJECTS/[name] before starting." Live work lives here. Stop pasting 15 context messages. Put them in files. 4. TEMPLATES/ ☑ Save your best work as reusable patterns: Prompt: "Use the template in TEMPLATES/[name]." Claude studies the structure before creating new. Add: "Match the length, sections, and tone. Only replace the content." 5. CLAUDE OUTPUTS/ ☑ The ONLY folder Claude writes to: One subfolder per project, mirroring PROJECTS/. You always know where deliverables live. ↳ No more "Untitled 47.docx" on your screen. 6. Global Instructions ☑ Set it once. It runs every single time: Settings → Cowork → Edit Global Instructions. Folder protocol, naming conventions, operating rules. You tell Claude how to behave ONCE. 7. The One Prompt ☑ Your new prompt is one sentence: "I want to [TASK] so that [SUCCESS CRITERIA]. Start by using AskUserQuestion." 80% of my chats start with exactly this. ↳ Claude prompts YOU. Not the other way around. 8. Obsidian (free) ☑ Open your Cowork folder as a knowledge base: Your .md files start looking like Google Docs. You can edit, search, and link them in seconds. Your brain + Claude's memory, in one clean place. 9. Alignment ☑ The folder does the talking. The prompt is a line. Claude asks YOU questions before doing anything. This replaces the old prompting era. I teach you how Claude works at claude-co.work. Copy my exact folder structure + download my personal .md files for Claude here: Step 1: Subscribe for free → how-to-ai.guide. Step 2: You will have two choices: free or paid. Step 3: Choose the free tier. Don't pay for anything. Step 4: Open your welcoming email (wait 30 sec). Step 5: Access my entire folder template & md files. Step 6: Send this image to your team's channel. Step 7: Read 2x newsletter per week (for free). Step 8: Become the "AI guy" at work, forever.

ANTHROPIC JUST PROVED MOST PEOPLE HAVE NO IDEA HOW TO PROMPT CLAUDE. Their applied AI team dropped a 24 minute free workshop. Not a creator who reverse engineered it. Not a Reddit thread. ANTHROPIC. The people who wrote the weights. And what they showed is uncomfortable. There are 6 elements to a properly structured Claude prompt. Most people are using 1. Maybe 2. That is not a skill issue. That is an information issue. And it has been quietly costing you every single day. The outputs that felt slightly off. The responses you had to rewrite 4 times. The prompts that worked once and never again. All of it traces back to the same 6 missing elements. The people who watch this 24 minute workshop tonight will understand something about Claude that most daily users still do not know exists. The people who skip it will keep getting 30% of what the tool is actually capable of and wonder why the results never quite land. I watched it twice. Then I built a Claude Skill that applies all 6 elements to every prompt automatically. No more thinking about structure. No more guessing what Claude needs. The framework runs in the background every single time. Full breakdown and skill setup is below. Bookmark this now. Watch the workshop first. Then read the guide. This is the one that compounds. Follow @cyrilXBT for the exact prompt architecture, Claude skills, and systems I use to get outputs most people do not believe came from one person working alone.

Roblox founder @DavidBaszucki bootstrapped his first company to a $20 million exit, then spent two years failing to find a CEO job before building Roblox in his early 40s — no revenue, no investors, pure vision. Today, Roblox has over 150 million daily users, 13 billion hours of monthly engagement, and a virtual economy worth over $40 billion. Here’s our conversation: 0:00 Roblox Origin Story 1:14 Sabbatical and Intuition 3:36 Founder vs CEO Mindset 5:43 Building the Clock 7:57 Lifestyle Startup Phase 8:49 First Product Failure 15:48 Buying First Users 17:43 Studio Goes Live 18:53 Roblox vs YouTube 21:59 Beyond Games Vision 25:50 Roblox Operating System 33:55 Nine Companies Inside 36:19 Safety and Monetization 41:13 Robux Economy Loop 45:19 Creator to Entrepreneur 45:49 Chasing Photoreal Concurrency 49:11 Imaginary Competitor Mindset 50:08 Capital Efficiency Playbook 52:11 Performance As Growth 55:40 Owning The Stack 58:36 Roblox Infrastructure Engine 1:02:32 Safety And AI Moat 1:06:57 Data Ethics And NPC Testing 1:11:31 Creator Earnings Explosion 1:16:08 Marketplace And Transparency 1:20:01 Near Death Lessons 1:24:43 Ads And Creator Discovery 1:25:35 Closing Reflections Includes paid partnerships.

This is the moment NVIDIA should be seriously worried. In the next couple of weeks DeepSeek V4 will be launched. It’s a direct attack on the entire AI stack that American companies have spent years locking down. Full “de-NVIDIA-ization”, a complete shift away from CUDA into Huawei’s CANN ecosystem, running on Huawei Ascend chips. That means one thing, breaking the dependency that made NVIDIA untouchable. 35x faster inference vs early versions. Nearly 3x the performance of NVIDIA’s H20 on a single card. 40% less energy consumption. Over 95% CUDA compatibility with migration times collapsing from months to hours. Even Jensen Huang has already admitted it. If this works at scale, it’s a “terrifying outcome” for US companies. Because here’s the real problem, this isn’t happening in isolation. Chinese tech giants like Alibaba, ByteDance, and Tencent are already ordering hundreds of thousands of Ascend chips. Market share is shifting fast, domestic chips now at 41%, NVIDIA slipping to 55% in China’s AI server market. Additionally DeepSeek V4 is reportedly offering API costs at a fraction of US competitors. $300 for massive workloads that would cost $2,500+ on OpenAI models, or even $5,000 on Anthropic. So this isn’t just about one model. It’s about China building a fully independent AI stack, chips, frameworks, models, and applications. Completely outside of US control. NVIDIA doesn’t just lose sales. It loses its grip on the global AI standard.

INSTEAD OF WATCHING NETFLIX TONIGHT. Spend 1 hour with this. Claude AI FULL COURSE that teaches you how to BUILD and AUTOMATE anything. The people who watch this tonight will wake up tomorrow with a skill that most people will not have in 2 years. The people who skip it will still be watching Netflix next year wondering why nothing in their life has changed. Your call.

LLM Knowledge Bases Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So: Data ingest: I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them. IDE: I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides). Q&A: Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale. Output: Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base. Linting: I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into. Extra tools: I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries. Further explorations: As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows. TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

Vibecoders right now:

🚨BREAKING: ANTHROPIC IS GIVING AWAY THE SAME CERTIFICATION THAT DELOITTE IS MASS-TRAINING 15,000 EMPLOYEES TO GET. It costs $0. You need a laptop. That's it. It's called the "Claude Certified Architect." Think of it like the AWS cert but for AI. If you were around when AWS certs started, you know what happened. They went from "cool to have" to "you're not getting hired without one." That took about 5 years. This is going to happen way faster. Look at who's already moving: Accenture - training 30,000 people on Claude Cognizant - rolled it out to 350,000 employees Deloitte - opened Claude access to 470,000 people Infosys - anchor partner These aren't startups experimenting. These are billion dollar consulting firms restructuring their entire workforce around Claude. And the certification they need? You can take it right now from your bedroom. Let me be real though. This is not one of those "watch 2 videos and get a badge" type certs that nobody respects. This thing is hard. 60 questions. 2 hours. Proctored. Webcam on. No breaks. No googling. They drop you into real scenarios like designing a customer support agent that handles refunds or setting up Claude in a CI/CD pipeline. The wrong answers look right on purpose. They're the exact mistakes real engineers make in production. 720 out of 1000 to pass. People who took it are saying the agentic architecture and multi-agent orchestration sections are brutal. Most of the exam is about building AI systems that actually work in the real world. Not prompting. Not chatting with Claude. Architecting production systems. All the prep? Free. Anthropic put out 13 courses on their Academy. No paywall. The cert itself is free for the first 5,000 people. After that $99 per attempt. How to get it: 1. Join the Claude Partner Network (free) → partnerportal.anthropic.com 2. Start the free prep courses → anthropic.com/learn 3. Register for the exam → anthropic.skilljar.com 4. Take the official practice exam 5. Book the real one when you're ready It launched 10 days ago. Almost nobody has it yet. That's the whole point. Get it before it becomes the thing everyone has.

Claude users right now:

Where has this hack been my whole life!?

@georgesttock NANO

Nano banana + MakeUGC + Veo3 = Ad Factory This agent creates 100s of ads every day - No actors - No product - No $10k/month agencies You're able to clone your comepitors ads with AI Paste there ad -> Pick an avatar and regenerate. Comment "NANO" and I'll send you the agent + the full playbook (must be following)

OpenAI fired this 23-year-old from their Superalignment team. But he turned his insider knowledge into a $1.5B fund that's outperforming Wall Street by 700% this year. He says maybe ~200 people in SF understand what's *actually* happening in AI right now. Here's his thesis: 🧵

скучал по твиттеру все это время

time to conquer X again #whatsup

Microsoft releases 18 courses for learning generative AI They are available free of charge and were just updated last week. Contents and link to access below 🧵