Efficiently Reconstructing Dynamic Scenes One D4RT at a Time https://t.co/KYRaZ2kNeo

discuss: https://t.co/ef6QGrJhOC

Small, specialized ML models are not dead... in fact, they have never been better at their specialized tasks. eg: Reflection Removal specialized model VS Nano Banano PRO prompted to (perfect result vs epic fail) https://t.co/wAn1BDlaQg

Introducing: the new llama-cli 🦙🦙 > Clean looking interface > Multimodal support > Conversation control via commands > Speculative decoding support > Jinja fully supported https://t.co/LHVULahExL

@soul_surfer78 You want to build a ChatGPT-like interface just for yourself or for serving customers? If the former, maybe check out my example here: https://t.co/Vw4c5Jd3j9 https://t.co/mVps0PMNhU

+1 more

You can simply select any part of Grokipedia and ask Grok to help you understand topics more deeply and explain them with greater clarity, making it easier to explore complex ideas, discover insights, and learn smarter • Highlight → ask Grok → get a clearer, deeper explanation instantly

ELON TURNED 𝕏 INTO THE LOUDEST TRUTH MACHINE ON EARTH - SOME ARE AFRAID Free speech got the volume knob and suddenly the comfort crowd started sweating. The Exposure: • Users challenge headlines in real time with clips data and receipts • Community Notes shreds weak narratives in public view • The EU fines land right when too many people stop buying the script When the truth gets loud, control gets nervous! Source: @elonmusk, @X

Grok is going viral on TikTok. It's also the #2 productivity app here in Belgium. The normies are starting to get it. https://t.co/8WXb8pppcD

Elon Musk “A fully reusable orbital rocket may be the hardest engineering problem of all time. SpaceX is tackling it with the most talented group.” https://t.co/6dXYxbExzD

🚨ELON MUSK: "President Trump is the funniest person I know in real life. He's got a great sense of humor. He's naturally funny. It's effortless. When he had Mamdani in the office and he told Mamdani to call him a fascist, that was awesome." 😂😂😂 https://t.co/E3eupsFskC

ELON MUSK: "American food is my favorite food. Cheeseburgers are my favorite. If I had to say, like, only one thing you can ever have for the rest of life, it would be a cheeseburger because cheeseburgers are amazing. It's a genius invention." https://t.co/MzJ0dxaVl4

ELON MUSK: “We kind of have a lot of Star things, you know, so we get Starlink, Starship. Starbase is probably the coolest place on Earth. It's only like three feet above sea level. So we built a gigantic rocket factory and two giant launch towers down by the river, literally within sight of the Rio Grande. And on an actual sandbar, kind of had to have an inspirational name and then we made it a city. So it's an incorporated city. Like legally a city. You don't hear about new cities being formed that often. This is my world. Yeah, I've gone from land to world.”

ELON MUSK: "I've been to Disneyworld 10 times. My favorite ride is Space Mountain. It needs an upgrade." 😂😂😂 https://t.co/Oi4GTuNdvT

ELON: “FEAR ISN’T THE MIND-KILLER, CONTEXT SWITCHING IS” Elon says the real productivity killer isn’t fear, it’s constantly jumping between tasks, and he's got a point. Running X, Tesla, SpaceX, Neuralink, and xAI, he describes his daily reality as “information triage,” fighting off inbox chaos and trying not to overwhelm his brain. “I have a lot of inbound communication, so it’s basically information triage. I try to segment my days to avoid too much context switching. Because arguably, fear isn’t the mind-killer, context switching is. It’s hard not to context-switch when your inbox is full of stuff, but think about it: if you had to context-switch every 3 seconds, every 30 seconds, or even every 3 minutes, the cognitive penalty would be enormous.” Source: @KatieMiller, @elonmusk

BREAKING: Elon Musk says modern fashion is stuck. Everything still looks like 2015 and we have not moved the needle in years. Past decades had clear styles while today suits barely change. We should evolve our look and bring back bolder, cooler ideas for a new vibe. https://t.co/m2Rxq9OBux

Awesome sunrise reflection at Starfactory https://t.co/XCwTkXdBRC

🚀 Google Antigravity, explained with no tech jargon Imagine telling AI: “Build me a full app while I drink coffee.” ☕🤖 Antigravity isn’t sci-fi – it’s Google’s new AI coding platform where: ✨ You describe what you want 🛠️ Multi-agent AI plans the architecture, writes code, tests in a real browser ✅ It shows you proof (screenshots, videos, task lists) 🚀 You approve → it deploys It can build: 🛒 E-commerce sites 📱 Mobile apps 🧩 Backend APIs 🐞 Bug fixes & code migration ⚡ MVPs in hours, not months In the old world, you = the entire dev team. In the new world, AI agents = your 24/7 dev team. 👉 Question: Would you trust AI agents like this to ship your next app? Reply: Yes / Not yet / Only for prototypes #GoogleAntigravity #AgenticAI #AIAgents #DevTools #FutureOfCoding

Claude Code can now run agents asynchronously. Huge for productivity. You can run many subagents in the background to explore your codebase. Work continues uninterrupted. When subagents complete tasks, they wake up/report to the main agent. Workflows feel faster already! https://t.co/WKppUrK1N3

Where will entry-level developers gain the experience they need, as AI takes over grunt work? At Open Source Summit Japan, Sakana AI researcher @Stefania_Druga suggests AI could help solve this conundrum. https://t.co/7YekTkMPO3 “AI may be getting smarter, but humans must not become less so. If the communities that underpin shared technical knowledge collapse—open source, mentoring networks, Q&A sites, classrooms—we will lose not only future developers, but the collective ability to understand and shape our tools.”

window seat reflection removal Reflection Removal through Efficient Adaptation of Diffusion Transformers https://t.co/siGFjHXojo

discuss: https://t.co/AKbOlqmQAI

app: https://t.co/cqeGDRutV0

Let's go! https://t.co/A2DwbXkVYP https://t.co/89FabZfv9l

Let's go! https://t.co/A2DwbXkVYP https://t.co/89FabZfv9l

+1 more

PaCoRe🔥 a new way to scale test-time compute, released by @StepFun_ai ✨ Breaks the context limit with parallel + coordinated reasoning ✨ Reaches million-token TTC without larger context windows ✨ Helps AI think in parallel, not just in long chains ✨ Models, data, inference pipeline all released on @huggingface

10 min to make an mcp tool for DuckDuckGo search in Reachy-Mini Reachy-Mini as a drop-in Alexa replacement. with atittude. and body. Nothing more than Alexa you could say... but also... I want it! h/t @ailozovskaya https://t.co/MKelPQvg1j

Featured HF Spaces this week = absolute bangers! 🔥 🔹 Ministral WebGPU: multimodal AI right in your browser 🔹 VibeVoice Realtime 0.5B: Microsoft fast TTS 🔹 LongCat Image Edit: SOTA text-guided image editing 🔹 Z-Image Turbo (ZIT) ControlNet: powerful guided generation https://t.co/xHcWGFE33M

🎨 Small QoL update for Trackio users: color-coded labels for each run that match the plot line colors https://t.co/Z1IqMmhd4m

🎨 Small QoL update for Trackio users: color-coded labels for each run that match the plot line colors https://t.co/Z1IqMmhd4m

MoE performance optimisation finally coming to transformers: https://t.co/2OQr7SzNQ7. Tell me this is not all in vain and moe are still the thing right? 🧐 https://t.co/DPUceOIq1E