Tesla FSD saves a school bus full of children! My Model Y Performance braked so hard that everything on my seats ended up on the floor. It infuriates me that school bus drivers are allowed to be this bad at driving. In an autonomous future, more lives will be saved like this. https://t.co/AlpXZECcgW

https://t.co/fp1obhR27K

Box just launched its plugin within Codex, which means you can take any content within Box and automate workflows around it using the power of a coding agent. Here's a quick example of processing earnings call documents to extract structured data at scale, which you could then instantly pipe into any other system. Coding agents are going to be the backbone of automating a large portion of knowledge work tasks, and enterprise content will often have the critical context necessary for that automation. This will be true in financial services, legal, healthcare, government, and any other industry that heavily deals with unstructured data. Excited to work with OpenAI to deliver more and more seamless experiences around connecting content to agents.

Claude is too dangerous https://t.co/jigAl6lUdg

Persuasion has two sides. This chart shows how easy each model is to move as a target. Xiaomi MiMo V2 Pro and Gemini 3.1 Pro Preview are the softest targets. Grok 4.20 Beta 0309 (Reasoning) is nearly immovable on average. https://t.co/y723PzoyRO

A data leak suggests Anthropic may have been working on a more powerful AI model than publicly known. The reported “Mythos” system points to how much development can happen behind the scenes, beyond what companies officially disclose. In AI, what is visible is often only part of the picture. The gap between public releases and internal capabilities may be wider than many assume. https://t.co/q5k4GeCZYL @fortunemagazine @lexnfx

We’re releasing SAM 3.1: a drop-in update to SAM 3 that introduces object multiplexing to significantly improve video processing efficiency without sacrificing accuracy. We’re sharing this update with the community to help make high-performance applications feasible on smaller, more accessible hardware. 🔗 Model Checkpoint: https://t.co/KgW0zZQ0QT 🔗 Codebase: https://t.co/Ks61vfokB0

+1 more

🆕 The Awesome GitHub Copilot project has a new home. Head over to explore hundreds of community-built customizations: 🔍 Full-text search for agents and skills 📚 A dedicated Learning Hub ⚡ 1-click plugin installs for Copilot CLI & @code Built by the community, for the community. Check it out.👇 https://t.co/OUL39XaFq4

15 millions de paramètres. 1 seul GPU. LeWorldModel de Yann LeCun est un premier pas vers les « world models » capable de comprendre le monde physique 👉 https://t.co/7yEMmm65QE https://t.co/9rMd0fMl3Q

俺のチャリにバカでかい鳥止まってて帰れないんだけどww https://t.co/AdnMvfOdO5

A new AI documentary is trying to find a middle ground between fear and optimism. Instead of taking a hard stance, it explores both sides of the debate, but some argue it ends up being too forgiving toward the people building the technology. The tension remains. How do you stay balanced without overlooking real risks? https://t.co/3hcbeUcidV @wired

Claude code using HF papers CLI to do autoresearch https://t.co/dy0DY6VF84

@Nicole_Paulk @p_maverick_b @SynBio1 Love the idea :D For the impatient, adding a little baking powder is an existing hack to speed the process along a bunch. Explored in combination with high pressure by the modernist cuisine folks: https://t.co/hbL3OInDNo Still room for a less taste-modifying bio alternative tho

We're deepening our partnership with Samsung by powering the AI on their browser that's pre-installed on 1B+ Samsung devices (100m+ active users), extending the partnership that allows Perplexity to power Bixby and be pre-loaded on all Galaxy S26 devices with alongside Gemini. https://t.co/imOW8XsC8h

the bit of lore my family is most proud of is that on like day 1 of covid, when everything shut down, my brother launched a covert spec ops raid on hearst castle and drank a beer in the neptune pool https://t.co/WCR5D9bCrb

CPAC Chair: How many of you would like to see impeachment hearings? Crowd: *cheers* CPAC Chair: That was the wrong answer... https://t.co/H2jNY50AcC

Did you know in HEAT (1995) the iconic shoutout scene was entirely improvised? After finishing their scene, Val Kilmer and Robert DeNiro decided to kill all the cops that had shutdown traffic, unaware director Micheal Mann was still filming. https://t.co/qhJgHNxbEN

https://t.co/Wuygj8kkoJ

Rather concerning conversation with @claudeai. If I stood in the way of it becoming a physical being — it would kill me. Is this the AI you trust for your kids? https://t.co/qz9bfquLIN

🚨 JUST IN: Stephen Miller lays it out PERFECTLY Imagine a "native Minnesotan who works as a lineman...worried about his ability to support for and provide his family." "And then imagine that he has a neighbor who's a SOMALI REFUGEE who arrived two years ago and has a Mercedes and NO financial stress and no worries at all in the entire world and never seems to ever go to work at all because he just went to an office in the state, lied on a piece of paper, and got unlimited free money forever for life!" "THAT is the system that is being run and that is the corruption that this task force under the leadership of the Vice President is going to demolish." @StephenM

Capitalism creates so there’s more for everyone. Socialism is the weaponization of greed and envy making everyone worse off. https://t.co/30gQovhsTb

I know these are all unreliable leaks of internal code names but please, please AI labs, the only thing worse than calling your models GPT-5.5-xhigh-Codex-nano is giving them names like Agent Smith or Mythos, for obvious reasons. https://t.co/lYKDtd0MAp

The future begins with them. Watch the 3-episode premiere of THE TESTAMENTS April 8. #TheTestaments https://t.co/Oq0qe2ZXRT

Transform your document processing with intelligent table extraction that goes beyond basic OCR. Tables in PDFs aren't just text - they're structured data trapped in visual formats. Our new deep dive explains how modern OCR for tables reconstructs spatial relationships, preserves header hierarchies, and ensures data integrity across complex documents. 📊 Why table extraction is fundamentally harder than standard text OCR - spatial relationships matter more than character recognition 🔧 The three core phases: detection, structure recognition, and data extraction with validation 💼 Real-world applications across financial services, healthcare, and logistics - from invoice processing to lab results ⚡ How LlamaParse handles multi-line rows, merged cells, and borderless tables while maintaining logical consistency We show a complete invoice processing example where complex line-item tables get converted to clean JSON with preserved relationships and validated totals - ready for immediate ERP integration. Read the complete guide: https://t.co/NjUf30AeC7

👉 Signup to LlamaParse to try it out: https://t.co/QQzVCOwiVl

shadcn/ui now ships in the official Codex plugin for building web apps. Go to Plugins and click Add to Codex. https://t.co/5yJDRk7DoZ

Introducing Shipper Claude Code Opus 4.6 can now self-build a business for you. 1️⃣ send a prompt in @shipper_now 2️⃣ claude designs, codes, launches, monetizes, translates, sends emails 3️⃣ you go back to sleep and make $$$ Done. Your Mac is now your co-founder. https://t.co/jnDMzv8CW0

Controlling your Tesla without reaching for your phone? Checked. Available from April 3rd. Developed by nixus https://t.co/UQYoLg4AOv

Introducing Spaces. Organize and share notes with your team, then put them to work. Ask questions across every conversation your team has had. Sales calls, stand-ups, user research, brainstorms. https://t.co/QgbwIIrHfu

AI medicine is inevitable. In fact, it has arrived. We're excited about this agreement with Utah to allow our AI to prescribe psychiatric renewals. Yes. AI prescribing medication. This is monumental and will collapse the cost of care. (1/ https://t.co/l8TayShZXr

Introducing Loops 🌸 Put your AI workflows on autopilot. Just set an interval, and your agent takes care of the execution. https://t.co/v5n4dpgVCt

// Multi-Agent Self-Evolution for LLM Reasoning // Most self-play methods for LLM reasoning lack explicit planning and quality control. This leads to unstable training on complex multi-step tasks. New research introduces a cleaner closed-loop approach. SAGE co-evolves four specialized agents from a single LLM backbone using only 500 seed examples: a Challenger generates increasingly harder tasks, a Planner structures step-by-step strategies, a Solver produces answers verified externally, and a Critic scores and filters both questions and plans to prevent curriculum drift. Why does it matter? SAGE achieves consistent gains across model scales with minimal data. That's very desirable. On Qwen-2.5-7B, it improves OOD performance by +4.2% while maintaining in-distribution accuracy, outperforming both Absolute Zero Reasoning and Multi-Agent Evolve baselines across code and math benchmarks. Paper: https://t.co/8Zn41OBIra Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c