Your curated collection of saved posts and media
Run remote multi-agents on blackbox Claude Codex, Codex CLI, Gemini CLI and Blackbox CLI competing on the same task—one subscription. https://t.co/p4owVFOeVx
#KostasKeynoteLessons: Curious about the "Keynote magic" behind my slides? I’m releasing the full Keynote source file for my recent Gaussian Splatting lecture, all 10 GIGAbytes of it! Grab the files in the thread and feel free to remix. https://t.co/2WeyXzB14J
Stream Ring - Save notes, control music, ask questions - No subscription req'd - Water resistant - Free sizing kit - Rechargeable https://t.co/6iBvHSqVzJ

saving this reaction image for anything surprising this holiday season https://t.co/otB9IqcgRD

Starlink's high-speed, low-latency internet is now available in São Tomé and Príncipe! 🛰️🇸🇹❤️ → https://t.co/ZVnndjUwoz https://t.co/Gu75sn0NJf
https://t.co/bFh7VsTHgD
Fact https://t.co/QFDraVouCK
Done with the semester! 🥳🥳🥳 We had an oral exam… where students were asking any question to Yann. It's always refreshing to have these dynamic sessions, where you get your intellectual curiosity satisfied on the spot. 🤓🤓🤓 And then, so many photos with the students! 🥹🥹🥹 https://t.co/1YyfSFdjMy
𝘛𝘖𝘒𝘠𝘖 𝘜𝘗𝘋𝘈𝘛𝘌𝘚 Tokyo-based Sakana AI is building tools that bring Edo-period literature and art into today’s world. Research scientist Tarin Clanuwat shares how Tokyo’s preserved Edo heritage nurtures the startup’s work. https://t.co/2a648m8sjE #TokyoUpdates #EdoTokyo
AI: Microsoft, Amazon bet big, but where does India stand in the global race? https://t.co/itelSP1ZjP @Nik_Inamdar @bbc
Welcome @lovart_ai 🤝 as the latest customer for the tldraw SDK! Lovart will be using tldraw's canvas to power their AI design agent experience. https://t.co/xpLOvkBh8B
Yesterday, we helped launch the Agentic AI Foundation, which includes MCP. Today, we're announcing fully-managed remote MCP servers for Google APIs. We're starting with @googlemaps, @googlecloud BigQuery, GKE, and GCE. https://t.co/1p5mU2rk3W
Tried the new Scouts (@yutori_ai), it’s a game-changer for designers/founders — it scours the web and sends v high quality intel. The UX is very impressive! https://t.co/z9JuAaMMEX
✨ The next generation of AI responses is here with Grounding Lite - a lightweight MCP server that connects your agent directly to live Google Maps Platform info using any LLM. Try it now ➡️ https://t.co/lMvgUmrrmb Give your agent the ability to: 🏠 Discover Places and access rich details 🚘 Calculate expected drive times ⛅ Fetch weather conditions 🗺️ Visualize details on 3D maps
Visionary: Your World Model Carrier on the Web! This open, web-native platform brings real-time Gaussian Splatting (3DGS, 4DGS, neural avatars) & mesh rendering directly to your browser with WebGPU & ONNX. Say goodbye to heavy installs & hello to "click-to-run" dynamic 3D content!
@katedeyneka It broke on my account and refused to finish. https://t.co/lMIFOLpLpG

Powered by the world’s most advanced broadband constellation, Starlink offers high-speed, low-latency internet in 150+ countries, territories and markets around the world - even in the most remote areas. Explore where Starlink is available → https://t.co/G1myzWJQ6j https://t.co/8idE4GXuCp

I’ve been through a lot worse than anything Donald Trump or Pete Hegseth can throw at me. And I'm not backing down, because this isn't about me — it's about you and every American's First Amendment right. https://t.co/x8hZrOrRKy
NVIDIA put me to the challenge to answer one simple question: can I teleoperate a robot arm in Isaac Sim to pick up oranges and place them on a plate... without a VR headset? What makes this possible: This is LeIsaac, @LightwheelAI's collaboration integrating @NVIDIARobotics' Isaac Sim with @huggingface's LeRobot. You can run this entire workflow in your browser with no local GPU or driver installation. Just spin up an NVIDIA Brev instance and you're collecting sim data within minutes. It is also fully optimized to run on @NVIDIAAIDev's DGX Spark! Try it yourself: https://t.co/0FzWqH8Gp5 #Robotics
US state attorneys-general demand better AI safeguards https://t.co/XFQHSrgXZv @CristinaCriddle @ft
AI companies want a new internet and they think they’ve found the key https://t.co/pfQqIo0i3S @verge
if you want to win you gotta be early once the gates to Hyperborea are open… … it might be to late Heil Borio Vrilly Niggas | @hyperborio ⚡️ - zyklon #Hyperborea #Agartha #Yakub #Hyperborio #Vril #Crypto #NFT https://t.co/lYEQELCTha

HyperMorning everyone,another day another development 🔥 We are thrilled to announce that USDH (@nativemarkets) is now fully integrated on https://t.co/jX3cJLO1wv! Key Highlights: Instant, Multi-Chain Swaps: You can now directly bridge into @nativemarkets (USDH) on Hyperliquid spot from EVM/Solana/Bitcoin with lightning speed! Benefit from direct seamless deposits, lightning-fast swaps, and full multi-chain compatibility powered by our AI-driven infrastructure, leveraging @nativemarkets innovative architecture. 🔗 Get Started Now: Trade $USDH now: [https://t.co/4TNIKaLtTn] Learn more about $USDH: [https://t.co/LzihTy70u3]

Former long-time @SpaceX employee Abhi Tripathi on a SpaceX IPO: "Once Musk realized Starlink satellites could be architected into a distributed network of data centers, the writing was on the wall. That is the moment an IPO suddenly came into play after being unlikely for so long. If you have followed Elon’s tactics, you know that once he commits to something, he leans fully into it. Much of the AI race comes down to amassing and deploying assets that work quicker than your competition. A large war chest resulting from an IPO will greatly help his cause and disadvantage all others.” Abhi worked at SpaceX for 10.5 years starting in 2010. Source: https://t.co/DnxeiWVdkD
Here's why I think SpaceX is going public soon. https://t.co/KEZIjhEsTH
@rrhoover I know right. Now…. if we are talking about early snapchat, he was a fan favorite and had the best cameos on Justin Kan’s Snapchat. @qasar https://t.co/85yE5vrXE3

🐟 Applied Team メンバー紹介 🐟 Sakana AIのApplied Teamで働くメンバーを週替わりで紹介します! 今週は、Software Engineerの珊瑚 彩主紀(さんご・みずき)さん(@mzking622)。 👨💻Sakana AI ジョインまでのキャリア/専門性 大学・大学院では機械学習や自然言語処理を専門に研究。 新卒で入社した株式会社メルカリでは、マルチモーダルデータを活用した機械学習による違反検知システムの開発に携わりました。 その後、米国スタートアップに機械学習エンジニアとして転職。e-commerceやVacation Rentalサービス向けのパーソナライズド検索・リコメンデーション・需要予測システムの構築をリードしました。 Sakana AIではSoftware Engineerとして働いています。 🎯Sakana AI でのミッションは? Applied Research Engineerが生み出した研究/検証結果を最大限の出力でお客様へ還元することが私のミッションです。そのために、AI技術とお客様の課題を深く理解し、最適化したUI/UXを備えた安定的なサービスを提供することを目指しています。 🔬今、注目の技術テーマは? AI Agentによるコーディング支援に強い関心があります。 Sakana AIのSoftware Engineerは、フロントエンド、バックエンド、インフラをすべてフルスタックに担当します。そのため、スタートアップならではのスピード感は保ちながらも、toBサービス特有の高いセキュリティと信頼性を確保することが求められます。 両方を実現するうえでは、AI Agentによるコーディング支援を最大限活用することが不可欠です。 もちろん、AI Agentが常に完璧なコードを書くわけではありません。だからこそ、AI Agentをどう使いこなすかがSoftware Engineerの生産性を高める大きなキーファクターの1つだと考えています。 🔥仕事へのこだわりは? Software Engineerとしてのバランス感覚です。 例えば、セキュリティ対策など最低限行うべきことは徹底しながらも、PoCや検証段階では、スピードを優先することもあります。その際は、コードが煩雑になっても、まずは「動くもの」を素早く作ることが重要です。 一方で、多くの顧客が利用するプロダクトとしてリリースする段階では話が別です。 時間をかけてでも、アーキテクチャ設計、コーディングレビューを徹底し、Unit testやIntegration testを行ったうえでリリースします。 プロダクトのフェーズやタスクの性質に応じて、何を優先し何を割り切るかを判断する——その柔軟さこそが、エンジニアとしての私の強みでありこだわりです。 Engineering like fish — agile, adaptive, and always moving forward. ⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯ Sakana AIでは、最先端のAIの社会実装に挑む、多数の職種のスタッフを採用中です! 募集中のポジションについては以下をご覧ください。 https://t.co/1q07mb3TzE

Google launches managed MCP servers that let AI agents simply plug into its tools https://t.co/aAS8OQxnjF @RebeccaBellan @techcrunch
McDonald's pulls AI-generated Christmas advert following backlash https://t.co/oQ7OUzPQJZ @Cressup @bbcnews
Fintech Revolut Offers Ex-Employees Share Buyback at 30% Discount https://t.co/36RgqAKrqv @ft @pymnts
GROK DID IN 1 YEAR WHAT MOST AIs WON’T DO IN 10 Grok started the year as an upgrade and ended it as an ecosystem. January brought Grok 3, new voices, Artifacts, Personas, and tools baked directly into X. By spring it could see, search, talk, and handle files. By summer it wasn’t just answering, it was creating, analyzing, and running your day. Grok Vision gave it sight. Grok Tasks made it productive. Grok Auto and Spreadsheet made it useful beyond chat. Then came Grok 4 with its own cloud, followed by Imagine for image generation, Code for development, and Background Thinking that never stops learning. Fall added finance, video tools, and transparency features. November unleashed Grok 4.1 and its “thinking” variant. December sealed it with Grok Trends and Grokipedia, its self-powered knowledge core. Every few weeks brought another leap, another expansion, another signal that Grok is evolving in real time. What took others years, it did in one. This isn’t a roadmap. It’s acceleration made visible. Source: @XFreeze, Financial Times, Tech Crunch