223: Caos no Moltbook, Claude Opus 4.6, GPT-5.3-Codex

Sexta-feira é dia de repercutir as principais notícias da semana, no mundo da IA. Vem ver quem participou desse papo:
• Marcus Mendes (threads.net/@mvcmendes), host sob controle
• Fabrício Carraro (twitter.com/fabriciocarraro), co-host sob controle, Program Manager da Alura, autor de IA (casadocodigo.com.br/products/livro-inteligenci…) e host do podcast Dev Sem Fronteiras (devsemfronteiras.tech)
Links:
• Imersão IA Sob Controle no Vale do Silício (pages.startse.com/imersao-vale-do-silicio-alura)
• IA Sob Controle 222: Como estamos usando o OpenClaw, e por que ele muda tudo? (iasobcontrole.tech/222-como-estamos-usando-o-o…)
• TUTORIAL: How to run Open Source LLMs on your PC without paying APIs (youtube.com/watch?v=pEPg7_j5Zcg&t=431s)
• Anthropic ironiza anúncios no ChatGPT em campanha do Super Bowl (youtube.com/watch?v=kQRu7DdTTVA&embeds_ref…)
• Sam Altman critica campanha da Anthropic (x.com/sama/status/2019139174339928189)
• Anthropic lança o Claude Opus 4.6 (anthropic.com/news/claude-opus-4-6)
• System Card do Claude Opus 4.6 (google.com/search?client=safari&rls=en&…)
• Anthropic libera agentes no Cowork (techcrunch.com/2026/01/30/anthropic-brings-age…)
• OpenAI lança o GPT-5.3-Codex (openai.com/index/introducing-gpt-5-3-codex)
• ChatGPT cai para 45% de share nos EUA (bigtechnology.com/p/new-data-openais-lead-is-c…)
• Prioridade ao ChatGPT leva pesquisadores a saírem da OpenAI (ft.com/content/e581b7a4-455c-48e6-a87c-c39bb9c…)
• OpenAI lança o Codex para macOS (zdnet.com/article/openai-codex-mac-app-free-tr…)
• OpenAI anuncia o Frontier (theverge.com/ai-artificial-intelligence/874258…)
• Jensen Huang critica falta de foco da OpenAI (reuters.com/business/nvidias-plan-invest-up-10…)
• Nvidia deve investir US$ 20B na OpenAI (bloomberg.com/news/articles/2026-02-03/nvidia-…)
• Gemini chega a 750M de usuários ativos mensais (techcrunch.com/2026/02/04/googles-gemini-app-h…)
• Memo interno da Meta cria hype sobre o Avocado (theinformation.com/articles/meta-memo-new-avoc…)
• Meta testa app Vibes (techcrunch.com/2026/02/05/meta-tests-a-standal…)
• Meta começa a associar uso de IA a relatório de desempenho (theinformation.com/articles/meta-officially-ti…)
• Google publica o PaperBanana (paperbanana.org)
• Kling AI chega à versão 3.0 (prnewswire.com/news-releases/kling-ai-launches…)
• Grok Imagine 1.0 é lançado com vídeos em 720p e áudio aprimorado (macmagazine.com.br/post/2026/02/02/grok-imagin…)
• CEO da ElevenLabs diz que voz é a próxima interface de IA (techcrunch.com/2026/02/05/elevenlabs-ceo-voice…)
• Texto do Marcus sobre o futuro de apps com voz (9to5mac.com/2026/02/04/the-future-of-apps-is-v…)
• ElevenLabs v3 ganha uso comercial (elevenlabs.io/blog/eleven-v3-is-now-generally-…)
• Chatterbox Turbo (github.com/resemble-ai/chatterbox)
• Apple investiga aceleração de geração de voz por texto (9to5mac.com/2026/02/02/new-apple-study-shows-h…)
• Amazon MGM Studio planeja usar IA para acelerar produções (reuters.com/business/media-telecom/amazon-plan…)
• Mistral lança o Voxtral Transcribe 2 (thedeepview.com/articles/mistral-surcharges-vo…)
• Z.AI lança GLM-OCR (x.com/Zai_org/status/2018520052941656385)
• Alibaba lança o Qwen3-Coder-Next (qwen.ai/blog?id=qwen3-coder-next)
• Samsung libera Perplexity nas TVs vendidas no Brasil (tecnoblog.net/noticias/samsung-libera-perplexi…)
• Perplexity libera o Deep Research Advanced (x.com/AravSrinivas/status/2019129261584752909)
• Perplexity anuncia o Model Council (x.com/AravSrinivas/status/2019444331006791847)
• EB-JEPA (x.com/BasileTerv987/status/2019071652546355544)
• Tencent Cloud, DigitalOcean, e Alibaba Cloud ganham suporte ao OpenClaw (theregister.com/2026/02/04/cloud_hosted_openclaw)
• Humanos se infiltram no Moltbook (theverge.com/ai-artificial-intelligence/872961…)
• Moltbook vazou credenciais (reuters.com/legal/litigation/moltbook-social-m…)
• ClawHub tem extensões maliciosas (tomshardware.com/tech-industry/cyber-security/…)
• Nvidia adia lançamento de GPU gamer (theinformation.com/articles/nvidia-delay-new-g…)
• Intel deve fazer GPUs (techcrunch.com/2026/02/03/intel-will-start-mak…)
• Xcode 26.3 ganha suporte ao Claude Agent e Codex (apple.com/newsroom/2026/02/xcode-26-point-3-un…)
• Microsoft integration Claude e Codex AI no GitHub, GitHub Mobile e VS Code (theverge.com/news/873665/github-claude-codex-a…)
• Microsoft anuncia o Publisher Content Marketplace (searchengineland.com/microsoft-launches-publis…)
• Yoshua Bengio lidera relatório sobre segurança da IA (internationalaisafetyreport.org/publication/20…)
• India zera impostos para atrair infra de IA (techcrunch.com/2026/02/01/india-offers-zero-ta…)
• Por dentro da aposta de Elon Musk para trazer usuários ao transformar o Grok em um gerador de pornografia (washingtonpost.com/technology/2026/02/02/elon-…)
• Suécia envolve IA em estudo com mamografias de 100.000 mulheres (thelancet.com/journals/lancet/article/PIIS0140…)02464-X/abstract)
• IA Sob Controle 202: A relação entre IA e radiologia, com Dr. Felipe Kitamura (iasobcontrole.tech/202-a-relacao-entre-ia-e-ra…)
• Toward Efficient Agents: Memory, Tool learning, and Planning (arxiv.org/abs/2601.14192)
• How AI assistance impacts the formation of coding skills (anthropic.com/research/AI-assistance-coding-sk…)
• D4RT: Teaching AI to see the world in four dimensions (deepmind.google/blog/d4rt-teaching-ai-to-see-t…)
• Teaching Models to Teach Themselves: Reasoning at the Edge of Learnability (arxiv.org/abs/2601.18778)
• Self-Distillation Enables Continual Learning (arxiv.org/abs/2601.19897)
• The Hot Mess of AI: How Does Misalignment Scale with Model Intelligence and Task Complexity? (alignment.anthropic.com/2026/hot-mess-of-ai)
• Evo-Memory: Benchmarking LLM Agent Test-time Learning with Self-Evolving Memory (arxiv.org/abs/2511.20857)
Quer ir com o Fabrício Carraro, o Marcus Mendes e o Paulo Silveira na “Imersão IA Sob Controle e Alura no Vale do Silício (pages.startse.com/imersao-vale-do-silicio-alura)“? Vagas limitadas, corra para reservar a sua!
TechGuide.sh (techguide.sh), um …