Your curated collection of saved posts and media
It's CVPR paper time (big computer vision conference). My AI found a ton this morning here on X (every link goes to the original post on X): https://t.co/kiuZ7QXLzb
🛒 Ever wished you had an AI grocery shopper to build your cart based on personal needs and tastes? That's the concept that won the students behind MagiCart the silver award in Google Maps Platform Awards' retail category. 🏅 The AI-powered shopping tool was created by Kelvin Mathew, Vaidic Soni, and Meidad Troper. The team designed MagiCart to help users save on groceries with consideration to factors such as proximity, budget, nutrition and more ➡️ https://t.co/DEkjk4o0b4
🌐 I've just released Sentence Transformers v5.4: we're going fully multimodal for embeddings & reranking! Also featuring a modular CrossEncoder, and automatic Flash Attention 2 input flattening. Highlights in 🧵 https://t.co/IDaRPVYc2g
Introducing our Claude Code Cheat Sheet. Keep track of all the latest Claude Code commands, shortcuts, and best practices. All in one place. Easy to navigate. https://t.co/esazLftGnv
@NoemiTitarenco Um, that no longer is true if you have the bleeding edge of AIs. Look at https://t.co/kiuZ7QXLzb Every word there is written by AI from Levangie Labs. https://t.co/xiuJ80Twa9 which is a new cognitive architecture that thinks and writes much more like a human does.
NEW paper from Microsoft Every agent benchmark has the same hidden problem: how do you know the agent actually succeeded? Microsoft researchers introduce the Universal Verifier, which discusses lessons learned from building best-in-class verifiers for web tasks. It's built on four principles: non-overlapping rubrics, separate process vs. outcome rewards, distinguishing controllable from uncontrollable failures, and divide-and-conquer context management across full screenshot trajectories. It reduces false positive rates to near zero, down from 45%+ (WebVoyager) and 22%+ (WebJudge). Without reliable verifiers, both benchmarks and training data are corrupted. One interesting finding is that an auto-research agent reached 70% of expert verifier quality in 5% of the time, but couldn't discover the structural design decisions that drove the biggest gains. Human expertise and automated optimization play complementary roles. Paper: https://t.co/fWhG9I8vPP Learn to build effective AI agents in our academy: https://t.co/1e8RZKs4uX
guess what @SherryYanJiang just said on NATIONAL RADIO!! so cool that the @aiDotEngineer singapore story is literally being talked about on-air rn ❤️🔥 @agrimsingh @swyx @ivanleomk @aimuggle https://t.co/orREah2VH5
Rembrandt-2, the next iteration to Rembrandt-1, is a powerful face image generation model, which is still under training, and here are some results with 15% of the total training steps. The results look awesome! I will share the source code for both Rembrandt-1 and Rembrandt-2 on Github, alongside the trained weights, as this is an open model. Please note that, this is not the full trained model and the images have artifacts, it's just 15% of the total training steps, so this is not a right time to judge. Rem2 is still undertraining.
https://t.co/8iMxyAAEg5
https://t.co/8iMxyAAEg5
1/n 🧵 Introducing Gaussian Wrapping — a principled framework for extracting high-quality meshes from 3DGS! 🚲 We recover thin structures, like bicycle spokes, where all prior methods fail. Follow the thread for a brief overview and links! https://t.co/T5XY1pKMvF
Introducing Neural Harmonic Textures: our new method for real-time novel view synthesis that outperforms all 3DGS and NeRF derivatives including (finally) ZipNeRF in terms of quality across all benchmarks. The code is released (Apache 2.0): (https://t.co/O0a2JyZ9Bu) 🧵 https://t.co/KYSxAU8ED3
Anthropic's Mythos does not appear to show any acceleration of ECI. After normalizing Anthropic's internal ECI with @EpochAIResearch 's public ECI, it's clear that the two metrics are extremely close, and that Mythos is pretty much on trend, just slightly above GPT 5.4. /1 https://t.co/kZXk5L4XpG
Germany-based Agile Robots has introduced its humanoid robot, Agile One, designed for industrial environments. It features a dexterous robotic hand and runs on a physical AI system trained using real factory data, allowing it to work and interact autonomously. The company is now aiming to move Agile One into full-scale production.
The AI Festival returns for the 4th year, bigger and better than ever. A celebration of the artist and creative minds leading this revolution. Submissions open and tickets on sale. NYC and LA. You really don’t want to miss this. https://t.co/UHmUkkkJL1
The Runway AI Festival returns this June to NY and LA to celebrate the world's best work leveraging AI across Film, Design, New Media, Fashion, Advertising and Gaming. Submissions are now open and tickets are on sale. Don't miss your opportunity to be a part of the year's biggest
AI Models Trained on Physics Are Changing Engineering https://t.co/VwPJWUpvYm
AI Models Trained on Physics Are Changing Engineering https://t.co/VwPJWUpvYm
Introducing @cursorbuddyapp -- inspired by clicky by @FarzaTV (thank you sir) and written from the ground up to be multi-platform ready, it provides a complete cursor agent that knows what you're doing, and comes as: Feature list now and soon - desktop - cli - web injection script (script include) - MCP - full voice - computer use - proper tooling - plugins and custom extensions and tools - themes - a playground - local options - detects screenshots - *walks you through learning - *time coded voice to text to get natural streaming - *multiple "canvas" views for interacting - *image support - *draws and annotates your screen - *interacts / available with agents to help out - *multiple "buddies" with characterisations and skills - *multi-player with other remote buddies* *roadmap will post the repo in this thread shortly, all free, all MIT license.
Here is a little torch.Tensor.__repr__ hack that allows you to dump custom tensor attributes in addition or instead the normal tensor's contents I found it very useful when I need to debug mismatching shapes and I didn't want to add more debug code https://t.co/wgG64DmxJH https://t.co/fuv10izKfI
// Scaling Coding Agents via Atomic Skills // Most coding agents train end-to-end on full tasks like resolving GitHub issues. But complex software engineering is really a composition of simpler skills, and training on the composite makes it hard to improve the parts. This research formalizes five atomic skills for coding agents: code localization, code editing, unit-test generation, issue reproduction, and code review. Instead of optimizing for a single composite task, the authors apply joint RL across all five skills simultaneously. The result: an 18.7% improvement that transfers to unseen composite tasks like bug-fixing and code refactoring without any task-specific training. Decomposition is a powerful scaling strategy. Rather than throwing bigger models at harder tasks, this work shows that systematically training the building blocks produces agents that generalize better. The atomic skills framework also gives you a clearer picture of where an agent is actually failing. Paper: https://t.co/7LhB8hcTcJ Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c
🚨 Hermes Agent is one of the most powerful open source AI agent frameworks alive. But... nobody knows how to use it. This just changed. Someone mapped the entire ecosystem. 568 GitHub stars in 48 hours. It's called hermes-agent-orange-book. A complete guide from zero to production. 80+ tools, skills, plugins, and integrations documented in one place. Open the repo. Follow the Orange Book chapter by chapter and deploy a fully autonomous agent in an afternoon. Every major AI lab has internal guides like this. Now you have one too.
Beyond language, toward Physical Intelligence. 🌍 Introducing #EXAONE 4.5, our first open-weight Vision Language Model. By integrating text and visual understanding, it’s built to solve real-world challenges. 📍Check out our blog: https://t.co/i82axJE1Kw 🔗Hugging Face:https://t.co/sdUr8knH0d #LG_AI_Researh #LG_AI #EXAONE #LG #AI #Multimadal #LLM #Physical_AI
"But here is what we found when we tested: We took the specific vulnerabilities Anthropic showcases in their announcement, isolated the relevant code, and ran them through small, cheap, open-weights models. Those models recovered much of the same analysis. Eight out of eight models detected Mythos's flagship FreeBSD exploit, including one with only 3.6 billion active parameters costing $0.11 per million tokens. A 5.1B-active open model recovered the core chain of the 27-year-old OpenBSD bug." https://t.co/yBTiiMq1Xy
The safest model format just became everyone’s shared standard.🤗 https://t.co/Gjw8G90K3a
The safest model format just became everyone’s shared standard.🤗 https://t.co/Gjw8G90K3a
Did the famous @huggingface HQ tour in Paris! Huge thanks to @xenovacom 😊 Such an amazing space! https://t.co/AGXLB9ew6h
Did the famous @huggingface HQ tour in Paris! Huge thanks to @xenovacom 😊 Such an amazing space! https://t.co/AGXLB9ew6h
@RemcoAI Yes, this will be forthcoming. We'll also post all of the session recordings on YouTube in the next couple of weeks. In the meantime, feel free to browse the sessions at: https://t.co/uvKLKMKdmg
Enterprise AI isn’t early anymore. ~87% of companies use AI. $200B+ spent annually. ~30% reach production. So we’re taking this into a build challenge at @ai_expo @TechEx_Event - putting real builds in front of those who decide what gets deployed. → https://t.co/nW0FGbGmO3. https://t.co/lXxon2Fm1X
T256S #AI #Thermal Camera with embedded real time Super-Resolution coming out! 2.4T NPU, real time 256x192 -> 640x480, Micro-scale Meticulous Detail~ Support standalone/phone plugin/PC/SBC multi-platform, support raw data read! Check it out: https://t.co/RaVePxFkjG https://t.co/j3t6Vamm8d
PyTorch Conference Europe 2026 was 🔥🔥🔥 What were your main takeaways? #PyTorchCon EU https://t.co/NhezYu51gW
New in Lyzr: Live Sources 🔄 Your knowledge base now syncs directly with SharePoint, public websites, and soon Google Drive. Connect once → always up to date → better RAG. No more manual uploads. Your AI agents deserve fresh context. https://t.co/DpXKTOiHmB https://t.co/BHxeI2y7Mr
