07-19-Daily AI Daily
YuanSi Tech Insight Daily - July 19, 2025
YuanSi Daily
AI News Roundup
Meta has assembled a 3400-person AI team, targeting AGI fundamental research, AI products, and Llama 5 R&D. Li Auto secured China's first in-car AI safety certification, with its in-car large model passing dual national standards. ChatGPT macOS app launched recording mode, transcribing up to 120 minutes of audio and extracting key information.
Outdoor SLAM technology saw a breakthrough; the S3PO-GS framework broke multiple records and solved the scale drift problem. AI creative engine Creati's annual revenue surpassed ten million dollars, successfully connecting influencers and businesses to quickly generate personalized ad videos.
Apple released its 2025 Foundation Model Technical Report, and OpenAI introduced ChatGPT Agent, capable of actively selecting tools to complete tasks.Today’s AI Buzz
Meta Just Assembled a Massive 3400-Person AI Super Squad! 🤩 Meta has put together a super artificial intelligence team of over 3400 employees, led by former Scale AI CEO Alexandr Wang. This team aims for “fewer people, more GPUs” and is divided into four groups: AGI Fundamental Research, AI Products (like Meta AI assistant, similar to ChatGPT), Fundamental AI Lab (led by Turing Award winner Yann LeCun), and Llama 5 R&D. Meta is actively poaching talent from other companies, though this move has raised a few eyebrows among its own staff.
Li Auto Nabs China’s First In-Car AI Safety Certification! 🏆 Li Auto just snagged China’s first-ever in-car AI safety certification! Their in-car large model aced both GB/T45654 and GB45438-2025 national standards, putting them at the forefront of AIGC content safety and identification in the country.
ChatGPT Recording Mode is Here: Your New Meeting Sidekick! 🎤 OpenAI’s ChatGPT macOS desktop app just rolled out a new recording mode (for Plus users)! It supports recordings up to 120 minutes, automatically generating transcripts, extracting key points, and even creating plans or code. Don’t worry, the original audio gets deleted after recording. Currently, it’s macOS only.
Outdoor SLAM Tech Just Got a Major Upgrade! 🗺️ S3PO-GS, an innovative framework from a Hong Kong University of Science and Technology (HKUST) team, has made a major breakthrough in outdoor SLAM (Simultaneous Localization and Mapping) technology! It tackles the tricky scale drift problem and has smashed multiple records across the Waymo, KITTI, and DL3DV outdoor benchmarks. This framework’s secret sauce lies in its self-consistent tracking module, dynamic mapping mechanism, and joint optimization architecture. 🔗 Project Repository
Creati, the AI Creative Engine, Just Raked in Over $10 Million Annually! 🤑 Creati, an AI engine focused on advertising creativity, has hit big with over ten million users and annual revenue exceeding ten million dollars in just one year! This engine uses AI models to transform viral influencer videos into customizable templates, helping businesses quickly churn out personalized ad videos. Creati’s success stems from its deep understanding of marketing and its knack for building a solid platform connecting influencers and businesses.
Meta’s New AI Structure Revealed: Are They Channeling ByteDance Vibes? 😲 Meta’s new AI organizational structure has been revealed, sparking questions if they’re taking a page from ByteDance’s book! Led by former Scale AI CEO Alexandr Wang, Meta has assembled a 3400+ strong AI team. This fresh structure boasts four key departments: AGI Fundamental Research, AI Product Team, Fundamental AI Lab (under Yann LeCun’s leadership), and the Llama 5 R&D team. This bold move by Meta clearly shows their serious focus on AGI (Artificial General Intelligence) and large language models.
Xiaohu.AI Daily Wrap-Up (July 15-17): Nine 🔥 AI Tool Updates! This Xiaohu.AI Daily Report Summary (July 15-17) shines a spotlight on nine major AI tool updates! We’re talking about Vidu’s open-source video generation tool, Moonshot AI’s Kimi K2 model, Runway’s Act-Two motion capture model, ChatGPT’s audio transcription feature, LTX Studio’s LTX-Video 13B video generation model, OpenAI’s ChatGPT Agent functionality, Suno v4.5+ music creation tool, and MirageLSD’s real-time AI video style transfer tool.
Google’s Veo3: A New AI Video King, But That Price Tag is Wild! 💸 Google’s Veo3 video generation model is now available to developers via the Gemini API, allowing them to whip up high-definition videos from text prompts. But here’s the kicker: it costs a whopping $0.75 per second! Talk about a high price tag.
Veo3’s Future: Can Tech Slash That Hefty Price Tag? 🤔 Veo3’s arrival marks a huge leap forward for AI video generation tech, but its steep cost is really holding back widespread adoption. Looking ahead, the big question is whether future technological advancements can bring those costs down. Stay tuned! 🔗 Markitdown 🔗 Open Deep Research
Segment Anything (SAM): Your Go-To Image Segmentation Superpower! ✨ Segment Anything (SAM), the incredible image segmentation model from Facebook Research, makes it a breeze to snip out any object you want from an image. 🔗 Project Repository
Hyprland: The Wayland Desktop Environment That’s Stealing the Show! 🌠 Hyprland is quickly becoming the new hotness in Wayland desktop environments! It’s super customizable, dynamically tiled, and honestly, just looks awesome. 🔗 Project Repository
Gitleaks: Your Personal Code Secret Service! 🕵️ Gitleaks acts like your trusty code security guard, automatically sniffing out and flagging sensitive info like keys, passwords, and other secrets lurking in your code. 🔗 Project Repository
Apple Drops 2025 Foundation Model Report & OpenAI Unleashes ChatGPT Agent! 🤯 Apple just dropped its 2025 Foundation Language Model Technical Report, showcasing two groundbreaking multilingual, multimodal foundation language models. Meanwhile, OpenAI unleashed its brand-new ChatGPT Agent, which can independently pick tools and tackle tasks on its own!
🔗 Report Link 
🔗 OpenAI IntroductionDocsGPT: Your Go-To Open-Source Q&A Guru! 📚 DocsGPT, an awesome open-source tool, aims to help users get reliable answers from their knowledge sources, sidestepping those pesky ‘hallucinations,’ and even supports private information retrieval. 🔗 DocsGPT Project
VisionThink: Smarter, Faster Visual Language Models! 🧠 VisionThink is a seriously efficient visual language model that dynamically adjusts image resolution based on task difficulty. This makes it way more flexible and effective than models stuck with fixed compression ratios. 🔗 Project Repository
The Imitation Game: LLMs Learning to Think Like Turing Machines! 💡 The TAIL method is shaking things up in the world of large language models (LLMs) by teaching them to think like Turing machines! This approach significantly boosts LLMs’ length generalization capabilities across a variety of tasks.
Buckle Up! AutoSteer Makes Autonomous Driving AI Safer. 🛣️ AutoSteer is here to buckle up AI with its new multi-modal large language model for autonomous driving safety! This awesome solution boosts the safety of MLLMs without needing to retrain the entire model.
AWS Unleashes Agentic AI Suite: Get Ready for Faster AI Agent Rollouts! 🚀 Amazon Web Services (AWS) just dropped a complete suite of Agentic AI solutions, accelerating the deployment of AI Agents! This full-package offering comes with four core pillars: model application capabilities, security and reliability, scalability, and deployment & production capabilities, plus the brand-new Agent development architecture, Amazon Bedrock AgentCore.
Open-Source AI Girlfriend Bella is Breaking the Internet! 💖 Bella, an open-source 3D AI girlfriend project created by user Jackywine, is setting the internet on fire with its stunning 3D modeling effects! 🔗 Project Repository
ACL 2025 Paper: Meet Evaluation Agent, Your AI Model’s Best Friend! 📊 Evaluation Agent, an awesome AI model evaluation tool presented in an ACL 2025 paper, is ready to be your go-to expert! It can whip up customized evaluation plans based on your needs and generate professional analysis reports. 🔗 Paper 🔗 Code 🔗 Website
Immigration Data Security: Back in the Hot Seat! 🚨 Immigration data security is back in the spotlight! The U.S. Immigration and Customs Enforcement (ICE) has gained access to sensitive medical data of tens of millions of Americans, sparking serious concerns about personal privacy and data security.
Hackathon Awards Ceremony: It’s Showtime! 🏆 The Hackathon Awards Ceremony is about to kick off! Bolt.new is inviting everyone to tune in and watch the live stream of the hackathon awards. ▶️ Video Demo
White House Teaming Up with PragerU on AI? Wild Rumors Flying! 😲 A Reddit post is stirring up buzz, suggesting the White House is teaming up with PragerU on an AI project to ‘beautify’ the images of the Founding Fathers. Say what?!
