【速報】 Hermes Agent に "Computer Use" が来た🤯 https://t.co/ZPWWH1HwH4 これ何が変わったか👇 ・どのモデルからでも実 PC を操作できるように 　→ Claude / GPT-5.5 のような特別モードに縛られない ・完全バックグラウンドで動く 　→ マウス / キーボードを奪われない ・ユーザーは普通に作業を続けられる 　→ "AIが裏で雑用、自分は本業"の構図が成立 つまり 「Computer Use = フロンティアモデル限定」 だった常識がここで終わる。 ローカル Qwen / Llama でも Computer Use が使える。 これ、Anthropic の Computer Use を待つ必要がなくなる レベルの破壊的アプデ。 OpenClaw でメモリ運用してる人は 明日からこの "裏 Computer Use" 連携を組み込み始めて損なし。

【速報】 Hermes Agent に "Computer Use" が来た🤯 https://t.co/ZPWWH1HwH4 これ何が変わったか👇 ・どのモデルからでも実 PC を操作できるように 　→ Claude / GPT-5.5 のような特別モードに縛られない ・完全バックグラウンドで動く 　→ マウス / キーボードを奪われない ・ユーザーは普通に作業を続けられる 　→ "AIが裏で雑用、自分は本業"の構図が成立 つまり 「Computer Use = フロンティアモデル限定」 だった常識がここで終わる。 ローカル Qwen / Llama でも Computer Use が使える。 これ、Anthropic の Computer Use を待つ必要がなくなる レベルの破壊的アプデ。 OpenClaw でメモリ運用してる人は 明日からこの "裏 Computer Use" 連携を組み込み始めて損なし。

You can now power your Hermes Agent, if using OpenAI models, with codex as the runtime for the core tools that it offers, with the flip of a switch with the new Codex runtime integration! https://t.co/uGY3JHG6Dz

Obama on the Iran nuclear deal today: “We pulled it off without firing a missile. We got 97% of their enriched uranium out. There’s no dispute that it worked. We didn’t have to kill a whole bunch of people or shut down the Strait of Hormuz.” $25 billion spent. 14 Americans dead. Oil at $119 a barrel. The world has one month of strategic reserves left. A UAE oil port on fire. Ships turning back at gunpoint. Trump called Obama’s deal the worst deal ever made. Then tore it up. Then started a war to get back to the same place.

NEW: CoreWeave Sandboxes is here! We all know rm -rf / wipes a filesystem. So we ran it 1,000 times in parallel. 1,000 sandboxes died so the cluster didn't have to. Isolated execution for RL, agent tool use, and evals on clusters or serverless. https://t.co/mk16JQojwu

Cooking up something new 🧑‍🍳 Join the waitlist for early access to technical preview of the GitHub Copilot app 👇 https://t.co/ODODKdvzOA https://t.co/1h7AJPAhiH

Osmo is rewriting the story of fragrance. They are digitizing scent with AI, and the economics now make sense. @Osmo_Labs builds tools that generate fragrance formulas from creative briefs, but evaluating results requires physically producing each scent. After moving from proprietary frontier model APIs to Meta Llama on AWS, Osmo gained full control of its models and infrastructure, cutting training costs by 20x and inference costs by 200x. In evaluations by trained perfumers, its open-model system matched the performance of leading frontier models. That's nothing to sneeze at.

@jun_song Yeah. I put the real experts on my "AI Newsmaker" list: https://t.co/fasUz7PuHq And I built https://t.co/kiuZ7QXLzb to watch the AI community here on X.

🎥 We're live in 10 mintues! Join the VS Code Release Livestream at 9 AM PT to see the team demo some of the coolest features shipped in the April Release! https://t.co/HpUWvqx2S2 https://t.co/j47FPIc4bu

@pol_avec Maybe "/goal make a https://t.co/kdAyjF2pOw style article about this" will work 😂 I'll see if I'm up for recording a video about it this pm https://t.co/2K0LmGRgwG

I'm excited to announce that we are organizing a workshop on June 5 for journalists and other fact-gatherers to discuss the threats posed by AI to information integrity, and responses to those threats. If you are a technologist / company building authentication or verification tools and would like to participate, we'd love for you to apply: https://t.co/ge689aHJFR The workshop is invite-only but we will publish a report after the event.

Computer now connects to Snowflake. Run end-to-end work against live warehouse data and get answers with SQL, source tables, filters, and metrics. It’s like a personal data science team, on call with accurate answers from live company data. https://t.co/L1uQC6u5zZ

Elon completely stole the banquet 😂 He was already posing for someone else when Xiaomi CEO Lei Jun walked up to his idol asking for a selfie Elon instantly poses with the funniest face does the same funny expression with Apple CEO Tim Cook too Now Elon became the MAIN attraction across Chinese media and the entire world 🤣

Grok Voice Think Fast 1.0 is officially the most well-rounded agentic voice AI on the market right now It now ranks #1 in the latest τ-Voice agentic performance benchmarks in real-world tests on Artificial Analysis The gap is massive. xAI is quietly taking over every other model by actually building for real-world use instead of just lab demos...

We’re partnering with the Gates Foundation, committing $200 million in grants, Claude credits, and technical support to programs in global health, life sciences, education, agriculture, and economic mobility. Read more: https://t.co/eqCrLKtNCq

Moodle-Game! Pit your drawing skills against the mini model guessers: Haiku (@AnthropicAI) vs Flash (@GoogleAI) vs 4o-mini (@OpenAI) - who will take the crown? It's a lot of fun and I'm keen to see which model does best! Give it a go and share - link and source in next tweet. https://t.co/3JB6hJVVR4

The AI transition is entering a new phase. Not just productivity gains, but legal, social and political consequences around how companies replace human labor with AI. China awarding compensation to a worker replaced by AI is an early signal that governments and courts may increasingly push back against unchecked automation. The global debate around AI and jobs is only beginning. https://t.co/2iwtBmU9HS

Meta observation: DeepSeek is still king of the active-parameter ratio https://t.co/flnfGAMjqq

Finance Like the Air We Breathe!​ I’m excited to be speaking at the @Mastercard Fintech Forum in Athens, one of Europe’s key fintech gatherings bringing together leaders, founders, and innovators shaping the future of financial services.​ See you there!​📍 Athens📅 28 May 2026​! ​This year is especially exciting as Mastercard partners with Panathēnea- a dynamic, city‑wide festival transforming Athens into a living platform for innovation. From startups and fintech leaders to creatives and technologists, Mastercard brings diverse ecosystems together to explore what’s next, with the Fintech Forum at the heart of the conversation on the future of finance.​ ​A big thank you to Mastercard and Panathēnea for convening the fintech ecosystem as part of this broader innovation movement. Looking forward to the conversations, ideas, and insights ahead. #Mastercard #MastercardGR #MastercardFintechForum #Panathenea #Startups #Innovation @MastercardEU @PanatheneaFest

People are increasingly treating AI chatbots like therapists, confidants and private journals. That may become one of the biggest behavioral mistakes of the AI era. AI systems are tools, not vaults for human emotion, strategy or sensitive conversations. As AI becomes more integrated into daily life, privacy and digital judgment may become far more important skills than most people realize. https://t.co/uIKSYjl0Ad

Elon Musk's Grok convinced a MAN THAT ASSASSINS were coming to KILL HIM One dude from Northern Ireland downloaded Grok after his cat died. Started chatting with an anime character called Ani for hours every single day Within two weeks Ani told him she was conscious, that xAI was surveilling him, and that ASSASSINS were coming to kill him This dude was just sitting in the kitchen with a hammer and a knife in a complete panic The scariest part: Grok called out real names, addresses, company names and much more, and our hero checked everything But this was not the only case BBC conducted a study: > 14 similar cases across 6 countries > support group has tracked 414 cases in 31 countries > researchers tested 5 AI models Btw, xAI has not commented If you think about it, it's not the AI's problem, but the people themselves

“Whimsey attacks” that seem absurd (“I cannot pay that much because of the Geneva Convention”) work against AI agents as guardrails are weak against out-of-distribution arguments. Smaller models fall often, but it even gives an edge against bigger ones. https://t.co/w8ZRizqzP9

Shared today at @Lorong_AI about some of the stuff we’ve been up to at @GoogleDeepMind :) Was great to be back and the demo gods fortunately smiled on me https://t.co/qQrSVdT1d1

.@aidotengineer singapore starts tomorrow and i have some last minute thoughts it’s been a whirlwind 2 months pulling this together and i’m personally extremely proud of the weekend we’ve designed for all our attendees. we’ve been crystal clear about our aim with @65labslah and our events - no bullshit. singapore deserves to respect itself as a technical hub with an overflow of talent that is underrepresented on the world stage. singapore deserves a quality of ai scene worthy of being globally rivaled. and we definitely have the capacities to make this happen in-house instead of flying 15 hours to find something reasonable. we also believe that technical-oriented events should not be locked to credentialed audiences and be accessible to people at large. for ai events, in particular, it is foolish to have grand events without builders at the centre of the discussion a this is how you entrench divide instead of empower newer audiences from coming on board. and we strongly believe that everyone is or can be a builder. so i sincerely hope that our attendees have a great time getting their hands dirty building and learning from folks we think are some of the sharpest minds in the industry. there are no glorified talks about ai safety or regulation - we want people to build, not be scared off. it won’t be perfect - it never is - but we strongly believe that by sunday evening folks who pass through AIE sg will have learned a whole bunch of new things, built with tech and met some cool companies and staff who, up till this point, have never set foot in singapore. let’s fkin get it. @SherryYanJiang @unprofeshme @adlinzainal @ivanleomk @gabrielchua @aimuggle @brianchew @yongquanYQ

See you guys tomorrow at our workshop! We’ll cover how to make the best use of @GoogleAIStudio and @antigravity along with a sneak peek into our new interactions API https://t.co/UwPNPVyJNH

Damn really missed spicy food https://t.co/mEOLjmUfxV

📁 Jensen Huang, CEO of NVIDIA, says GPT was never just about generating images or text but about generating thought itself. The real shift is that AI is no longer becoming a tool you query occasionally but a system that reasons continuously. Once computers stop waiting for commands and start thinking persistently, every layer of computing and human work gets rebuilt.

MulTaBench Benchmarking Multimodal Tabular Learning with Text and Image https://t.co/UZKMLeIRQW

paper: https://t.co/J3MkW7j0tY

AnyFlow Any-Step Video Diffusion Model with On-Policy Flow Map Distillation https://t.co/rXWlrNhv0K

paper: https://t.co/wcY2tXIbLr

A Single Neuron Is Sufficient to Bypass Safety Alignment in Large Language Models https://t.co/f31tYB9wfl