Santuba

@andrebrandao_rj Chega de privilégio mal administrados , isso é crime

R$6 Mil reais de gasolina por mês é normal um vereador gastar esse valor ? Deixa a sua opinião nos comentários.

@RobertoReis @RenanSantosMBL ele é bom demais não tem jeito

Golaço do @RenanSantosMBL aqui.

desfavelização ja

It's simply undeniable just how much the U.S. regulatory environment has shifted in Chainlink's favor lately: ✅ U.S. SEC and CFTC issue joint statement classifying $LINK as a digital commodity ✅ @Grayscale LINK ETF launched (1.2% of circ supply accumulated) ✅ @Bitwise LINK ETF launched (0.27% of circ supply accumulated) ✅ @CMEGroup LINK futures launched on largest CFTC regulated derivatives exchange ✅ Sergey Nazarov appointed to @CFTC Innovation Advisory Committee, launched by @MichaelSelig ✅ Former Chainlink Labs Deputy General Counsel appointed to @SECGov Crypto Task Force, led by @HesterPeirce ✅ Chainlink and CCIP highlighted prominently in White House digital asset report ✅ Sergey and CLL regularly in D.C. meeting with U.S. regulators, legislators, & policy makers ✅ Chainlink chosen by U.S. Department of Commerce (@CommerceGov) to bring macroeconomic data onchain ✅ @Anchorage Digital and Chainlink join Blockchain Leadership Fund as founding members (new crypto PAC) ✅ Chainlink joins @BlockchainAssn & @DigitalChamber lobbying groups ✅ Sergey gives public remarks at White House Digital Asset Summit ✅ SEC issues interpretive guidance based technical recommendations provided during six in-depth briefings with Chainlink This opportunity is being very much seized! The sum is much greater than its parts, now let's let Clarity passed

Introducing Claude Design by Anthropic Labs: make prototypes, slides, and one-pagers by talking to Claude. Powered by Claude Opus 4.7, our most capable vision model. Available in research preview on the Pro, Max, Team, and Enterprise plans, rolling out throughout the day.

@Grayscale LINK. Todo dinheiro que entra eu quero comprar mais. virou um vicio

You don't own enough ___________.

LINK Everything

@chainlink vai começar mostrar todo seu poder acumulado. Esteja preparado

This 25-minute Claude Code workshop by Anthropic's own applied AI team will teach you more about Claude Code best practices and making your AI tools actually work together than everything you've scrolled past this year. Bookmark this & watch, no matter what. Then read the guide below.

@0xCVYH major qual o mac mini que voce usa?

passei a semana validando se SAE features podem ser usadas como reward signal em RL de reasoning — e o achado contra-intuitivo bateu: Stage Gate 1: pre-teste de correlacao em 100 rollouts GSM8K com Qwen3.5-4B. • SAE features (top-10 helpful - top-10 harmful, layer 18): ρ = +0.540 • Raw contrastive direction (layer 13): ρ = +0.508 ambos p<0.0001. sinal correlacional forte nas duas lados. Stage Gate 2: GRPO 100 steps, ablation matrix com 3 reward variants: • R0 outcome-only: 74% • R1 outcome + SAE features (λ=0.1): 76% (+2pp, convergiu 2.5x mais rapido) • R2 outcome + raw direction (λ=0.1): 65% (-9pp, harmful) o gap de 11pp entre R1 e R2 e o finding principal. mesma direcao causal. ρ 0.508 vs 0.540 (quase igual correlacional). mas em RL: SAE filtra o sinal, direcao raw injeta polissemia. GRPO amplifica ruido. SAE decomposition NAO e cosmetico. e o passo causal que converte sinal correlacional em sinal treinavel. zero SAE publicados em hybrid GDN (Qwen3.5) ou hybrid MoE (Gemma 4) ate isso. nossas SAEs sao as 2 primeiras publicas: huggingface.co/caiovicentino1… huggingface.co/caiovicentino1… paper em construcao. next: Stage Gate 3 full RL (2000 steps, per-token mech-reward) pra quebrar o ceiling 76% e bater os landmarks dense (AIRI +13.4% AIME).

Most people are still “prompting” Claude. Power users are building Skills stacks that turn Claude into an AI operating system. Here are 17 FREE Claude Skills resources (this is the starter pack everyone will be using soon): 👇 OFFICIAL DOCS • Best Practices — lnkd.in/emxu8Vsr • Skills Documentation — lnkd.in/eSzfnUNc • API Reference — lnkd.in/erjGW9q5 • MCP Documentation — lnkd.in/ejKJuNEX BLOG POSTS • Introducing Agent Skills — lnkd.in/enrM2tWr • Engineering Blog — lnkd.in/eRn5aYyQ • Skills Explained — lnkd.in/e8zEX2Fe • How to Create Skills — lnkd.in/eDaug-WJ • Skills for Claude Code — lnkd.in/eQpjSyBW • Frontend Design Skills — lnkd.in/efPCkgWb EXAMPLE SKILLS • Anthropic Official Library — lnkd.in/er2tG4ZB • Partner Skills Directory — lnkd.in/ejUcTPjT COMMUNITY LIBRARIES • skills.sh • skillsmp.com • smithery.ai/skills • skillhub.club If you're serious about: • vibe coding • shipping faster • building agents • making money with AI You need this. Bookmark now. Retweet to save for later. The people who build Skills early will dominate distribution.

In 14 minutes, this Anthropic engineer who wrote "Building Effective Agents" will teach you more about building them right than most developers figure out on their own in months. Bookmark this for the weekend. Then read the builder's guide below.

We’re building a latency Football bot for Polymarket. Target: operational before the World Cup. Save this. The principle: pro sports feeds (Sportradar, Opta) deliver pitch events in ~200-500ms. The Polymarket orderbook takes longer to reprice thin liquidity, market makers pulling quotes while they reassess. That window is the edge. Yesterday, first live test on Bayern vs Real (Champions League QF). 86th minute, Camavinga gets his second yellow. Pipeline receives the event from Sportradar 280ms after the card. “Bayern advances” market was at ~0.55. Model recomputes to ~0.68 post-red and fires a $500 order. Partial fill as expected: ~$180 caught around 0.55-0.57, the rest slipped to 0.63. Average 0.59. Market stabilized at 0.67 a few seconds later. Unrealized +$30. +6% in seconds. It’s a test. But the loop worked end to end detection, decision, fill, before the book caught up. What we learned: network latency is part of the problem. The real bottleneck is orderbook depth. We’re competing with sharp bots, not retail on their couch. And “next goal” markets have better spreads than qualification markets. Pivoting there. What we’re building before June: fill routing across 12 venues via Jito atomic bundles. Low-signal event modeling (dangerous fouls, injuries, tactical shifts). UMA oracle hedging. Node co-location near Sportradar servers. Why the World Cup matters. 104 matches in 39 days. $2.5B+ in projected prediction market volume. Deep liquidity means bigger positions fill cleanly. Thin liquidity in group stages means wider spreads. Both environments leave serious money on the table. 56 days to ship. We’re on it.

Permita-se imaginar

Introducing Claude Opus 4.7, our most capable Opus model yet. It handles long-running tasks with more rigor, follows instructions more precisely, and verifies its own outputs before reporting back. You can hand off your hardest work with less supervision.

3-HOUR ADVANCED CLAUDE CODE LECTURE FROM NICK SARAEV. SAVE THIS if you already use Claude Code daily - there's still new stuff in here > CLAUDE.md optimization, agent harnesses, task parallelization > Karpathy's autoresearch approach > browser automation - Computer Use vs Browser Use > workspace org, security, auto-mode, OAuth > where Claude Code is going watch it this weekend👇

> do you understand what just happened to the job market > one person + 67 Claude Skills > does the work of an entire dev team > $20/month vs $200K/year in salaries > the Google engineer who automated 80% of his job? > this is the exact skill list he used > bookmark this. seriously. right now. 👇

@0xCVYH Entendi foi nada

Hermes Agent (Nous Research) adicionou skill nativa: /architecture-diagram — agora built in depois do autor da skill original ter released em MIT license. o pattern e importante: agentes open source absorvendo skills de terceiros via portabilidade (MIT → built-in) em vez de depender de plugin marketplace proprietario. economy de skill modular, zero friccao de install. e exatamente a direcao pra onde 'agent runtime' ta indo quando nao esta refem de store. via @Teknium1: x.com/Teknium1/statu…

Dois engenheiros da Anthropic acabaram de mudar a forma como devs pensam sobre IA. Barry Zhang e Mahesh Murag subiram no palco do AI Engineer Code Summit e disseram uma frase que incomodou muita gente: "Parem de construir agentes. Construam Skills." Em 16 minutos eles provam que a indústria inteira está resolvendo o problema errado. Aqui está o que a maioria não entendeu: → Skills são pastas. Literalmente pastas com arquivos markdown. → Elas ensinam ao Claude o SEU fluxo de trabalho, a SUA expertise, o SEU domínio. → Um único agente genérico + biblioteca de Skills específicas supera dezenas de agentes especializados. → Fortune 100s já estão deployando Skills em escala pra ensinar agentes sobre processos internos. → Times de produtividade com 10.000+ devs usam Skills pra padronizar como código é escrito. A analogia que eles usaram é perfeita: Quem você quer fazendo seu imposto de renda? O gênio com QI 300 que nunca viu legislação tributária, ou o contador experiente que faz isso há 20 anos? Inteligência sem expertise é entretenimento. Expertise empacotada é produtividade. O que mudou: a Anthropic parou de tentar criar agentes diferentes pra cada domínio. Perceberam que com Claude Code, o padrão é sempre o mesmo. Um modelo acoplado a um runtime com filesystem. A diferença entre um agente medíocre e um extraordinário não é o modelo. É o conhecimento de domínio que você alimenta. Skills resolvem isso com progressive disclosure. O agente só carrega o nome e descrição da skill. Quando relevante, puxa o SKILL.md. Quando precisa de mais, navega os arquivos de referência. Zero desperdício de contexto. Isso não é uma feature. É uma mudança de paradigma. Quem entender isso agora vai operar em outro nível daqui a 90 dias. Quem ignorar vai continuar escrevendo prompts de mil palavras toda vez que abrir o chat. E ainda explicar de novo e de novo o que “realmente” quer.