The AI frontier is expanding into the physical world, accelerating toward superintelligence, and empowering developers at every layer of the stack. XPENG pushes embodied AI into real-world motion, OpenAI lays the groundwork for safe, society-scale intelligence, and GitHub arms builders with next-generation tools to turn ideas into production-ready systems.

🤖 XPENG Physical AI: Vision-to-Action Intelligence – XPENG unveils VLA 2.0, a vision-to-action model driving AI cars, humanoids, and flying vehicles. With Robotaxi, the IRON robot, and new flying systems nearing mass production, XPENG pushes toward lifelike robotics and global open-source expansion.

🚀 OpenAI 2025 Vision: Preparing for Superintelligence – OpenAI highlights rapid progress as AI surpasses top human benchmarks. With falling intelligence costs and major breakthroughs ahead, the focus stays on safe deployment, shared global standards, and broad access to AI-driven scientific and economic benefits.

💻 GitHub AI Stack: Empowering Developers – GitHub delivers an AI-powered dev platform with GPT-5 models, automated security, and end-to-end DevOps. Enterprises gain faster coding, stronger governance, and seamless AI agent management across industries.

From embodied intelligence and safe superintelligence planning to AI-native development tooling, the next wave of AI is not just smarter, it’s physical, responsible, and built for builders.

XPENG Unveils Physical AI Future

XPENG, now calling itself a "global embodied intelligence company," announced major advances at its 2025 AI Day, including the VLA 2.0 large model that instantly maps vision to action for AI cars, robots, and flying vehicles—bypassing traditional architectures. The event saw launches of a new Robotaxi, the Next-Gen IRON humanoid robot, and two flying systems, all built on XPENG's full-stack self-developed physical AI and ultra-scale cloud training. XPENG also revealed open-sourcing plans for VLA 2.0, global partnerships, and imminent mass production for both Robotaxi and flying cars, while highlighting its goal of achieving life-like, rapidly commercialized robots and expanding three-dimensional mobility experiences by 2026.

Preparing for Superintelligence: The 2025 AI Vision

OpenAI's 2025 report highlights rapid AI progress with systems now outperforming top humans in many challenges and embedding AI deeply into daily life, from healthcare to autonomous vehicles. The cost of intelligence is falling steeply, and AI is expected to make significant discoveries soon, with more accessible, empowering tools reaching widespread use. Safety remains a core focus, emphasizing empirical research, shared global standards, and regulatory innovation to mitigate risks of advanced AI. The vision is broad benefit through responsible deployment, enhanced scientific and medical breakthroughs, and new economic models supporting human flourishing in a world transformed by AI advancements. This evolving AI ecosystem is shaped by increased investment, improved benchmarks, and stronger public oversight to ensure the technology's positive impact grows broadly and equitably.

Empower Developers with GitHub AI Tools

GitHub offers a comprehensive AI-powered developer platform designed for enterprises and teams of all sizes, enabling faster, smarter, and more secure software development. With integration of cutting-edge AI models like GPT-5 and GPT-5 Mini, GitHub enhances productivity with AI-assisted coding, advanced security features including automated vulnerability scanning and compliance controls, and end-to-end DevOps workflows. This platform supports industry-specific solutions for healthcare, financial services, manufacturing, and more, facilitating seamless collaboration, governance, and automation. Trusted by top global organizations, GitHub combines scalable infrastructure with AI-powered tools such as Copilot for business, unified management of AI agents, and ecosystem integrations to empower developers from startups to Fortune 100 enterprises in delivering innovative, secure software efficiently.​

Hand Picked Video

n this video, we’ll look at Olly’s biggest update yet, Auto Commenter now supports Facebook, Instagram, and X! Discover how you can automatically engage with 60+ comments per hour, grow faster across platforms, and save hours of manual work every single day. If you're looking to scale your engagement effortlessly, this update changes the game.

Top AI Products from this week

• Sheet0 - YOLO mode comes to spreadsheets! 💪🏻 Ask a question, list your sources, and let hit auto-run: sheet0’s cloud browser grabs data in parallel, cleans it, and produces an interactive, tidy sheet for your enjoyment. Flip between charts, tables, and SQL. Share or export in one click.

• Skarbe - Skarbe is your inbox. Inside lives Oskar, your AI assistant who never forgets. While you handle emails, he works in the background: tracking conversations, following up on cold leads, recording calls, and organizing deals. You get back 2 hours a day.

• TheySaid 3.0 - TheySaid 3.0 is here, smarter, more powerful, and built for scale. Every one of your most requested features is in this release: 2-way voice, usability testing, AI forms, AI Insights 2.0, integrations, panel recruiting, and more.

• Lumi - Lumi is an AI reading experiment from Google's PAIR. Instead of summarizing, it adds a lightweight AI layer directly on top of arXiv papers. It highlights main points in the source text, makes references clickable, and lets you ask contextual questions about any selected text or image.

• Primer - Primer agents are experts on your product and give live, personalized walkthroughs like your best team member – from sales to support – but in any language, 24/7. We founded Primer on a simple belief: every product should come with an expert on tap.

• ComplyDo - ComplyDo automates how enterprises comply with hundreds of global regulations. Already used by global enterprises, ComplyDo helps organizations stay compliant continuously and scale across markets. Without a fleet of consultants.

This week in AI

• One ruler to measure them all - ONERULER benchmark tests multilingual long-context LLMs across 26 languages, revealing Polish tops over English and cross-lingual effects on performance.

• Comet Android Early Invites - Comet Android invites are live! Boost your early access chances with active Perplexity Android use and Pro/Max status. Shape the future of mobile AI browsing. More invites soon!

• Qwen-Image-Edit-2509 - AI-powered multi-image editing with enhanced consistency for portraits, products, and texts; supports ControlNet for precise, seamless light restoration.

• Google Gemma 3 (1B) - Google's lightweight multimodal AI model supporting 140+ languages and a 32K token context window for advanced text and image understanding in compact form.

• Real-Time Reasoning Agents - Real-Time Reasoning Gym tests agents in dynamic games, balancing fast reactions and long-term planning. AgileThinker excels by combining both reasoning modes for superior real-time decisions.

Paper of The Day

VisualQuality-R1 is a novel no-reference image quality assessment model trained with reinforcement learning to rank, capturing the relative nature of visual quality. It uses group relative policy optimization and the Thurstone model to compute multiple quality scores per image, providing precise and continuous fidelity measures. The model outperforms existing deep learning-based IQA methods, generates human-aligned quality descriptions, and supports multi-dataset training without perceptual scale adjustments. It excels across diverse distortions, enhancing interpretability and practical applications like post-processing guidance and quality-aware retrieval, though it faces challenges like computational cost and fixed prompt reliance.

To read the whole paper 👉️ here