Global AI Weekly
Issue number: 119 | Tuesday, October 14, 2025
Highlights
Meet Sora 2: AI Video Just Leveled Up
Say hello to Sora 2, OpenAI’s groundbreaking video + audio generation model with realistic physics, synchronized dialogue, and powerful controls. It even lets you drop yourself (via cameos) into custom video scenes. Download the new app, explore the feed, and see how we’re building AI that’s as fun as it is forward-thinking.
openai.com
A new AI strategy by the EU to cut reliance on the US and China
The EU's Apply AI strategy, backed by €1 billion in funding, seeks to establish artificial intelligence as a cornerstone for Europe's security and global competitiveness. This initiative aims to reduce dependence on AI technologies from the US and China, focusing on fostering innovation and strengthening Europe's digital autonomy. The strategy underscores AI's critical role in shaping Europe's future in a rapidly evolving technological landscape.
dig.watch
AMD wins massive AI chip deal from OpenAI with stock sweetener
AMD has struck a groundbreaking deal with OpenAI, securing a major AI chip partnership. As part of the agreement, OpenAI will have the option to purchase a 10% stake in AMD at just a penny per share, adding a unique twist to the collaboration. This move positions AMD as a key player in the competitive AI chip market while strengthening ties between the two tech giants.
arstechnica.comResearch
QUASAR: Quantum Assembly Code Generation Using Tool-Augmented LLMs via Agentic RL
This paper introduces QUASAR, a novel approach to generating quantum assembly code by leveraging large language models (LLMs) augmented with specialized tools and guided by agentic reinforcement learning (RL). QUASAR is designed to tackle the unique challenges of quantum programming by combining the strengths of LLMs with task-specific enhancements, enabling efficient and accurate code generation. The approach paves the way for advancements in quantum computing workflows through intelligent automation.
arxiv.org
How to Teach Large Multimodal Models New Skills
This paper introduces a method for teaching large multimodal models new skills without retraining them from scratch. By leveraging a technique called progressive learning with examples, the approach allows models to adapt effectively to new tasks and domains. The research demonstrates how this strategy enhances flexibility and efficiency in expanding the capabilities of these advanced models while preserving their previously learned knowledge.
arxiv.orgVideo
AI in Fintech: From Invoicing to Intelligent Payment Systems
AI is revolutionizing financial software, transforming traditional processes like invoicing into intelligent, machine learning-driven payment platforms. This discussion highlights the journey of developing scalable AI-powered financial tools that optimize payments and streamline operations for millions of businesses. With advancements in predictive systems, payment optimization, and natural language processing, AI is shaping the future of fintech architecture and smart financial solutions.
youtube.com
OpenAI DevDay 2025: Opening Keynote with Sam Altman
Sam Altman kicks off OpenAI DevDay 2025 with an inspiring keynote, sharing groundbreaking ideas to transform how developers approach building with AI. The session features exciting announcements, live demonstrations, and a forward-thinking vision of AI's role in shaping the future of development. It offers a glimpse into the innovative possibilities AI brings to the tech world.
youtube.comArticles
Effective context engineering for AI agents
Anthropic shares insights on effective context engineering for AI agents, focusing on building systems that are reliable, interpretable, and adaptable. The content explores techniques to design AI that can better understand and respond to context, enhancing both performance and safety. It emphasizes the importance of aligning AI systems with human needs while maintaining clarity in their operation.
anthropic.com
OpenAI wasn’t expecting Sora’s copyright drama
OpenAI adjusted the copyright settings for its Sora generator and introduced more options for users to manage their likenesses in response to early feedback. These changes come after concerns were raised about how the tool handled intellectual property and personal image use. The update highlights OpenAI’s efforts to address user concerns and improve transparency in its AI tools.
theverge.com
TX-GAIN: MIT supercomputer to power generative AI breakthroughs
MIT's Lincoln Lab has introduced TX-GAIN, the most powerful AI supercomputer at any U.S. university, designed to accelerate advancements in generative AI. This cutting-edge system will support groundbreaking scientific research, offering unmatched computational power to tackle complex challenges. It represents a significant leap in AI capabilities for academic institutions nationwide.
interestingengineering.com
Will AI ever win its own Nobel? Some predict a prize-worthy science discovery soon
The possibility of autonomous AI scientists making groundbreaking discoveries sparks debate among experts, with questions about their feasibility and desirability. Some predict AI-driven innovations could soon lead to Nobel Prize-level achievements, potentially transforming the landscape of scientific research. However, concerns linger over the role of human intuition and creativity in such advancements.
nature.comUpcoming Events
AgentCon 2025
After an exciting set of AgentCon events before the summer holidays, AgentCon is happening in more locations during the second half of the year. The AI Agents World Tour is a series of one-day conferences for developers focused on autonomous AI agents. Spanning cities like San Francisco and Singapore, these events gather top engineers, researchers, and creators shaping the future of AI technologies. It’s an opportunity to connect, learn, and innovate in the fast-evolving world of autonomous AI. Check all events and register for an event near you!
globalai.communityCode
ChatGPT Just Got an Upgrade: Introducing Apps
ChatGPT now supports apps, bringing extra power, plugins, and custom tools right inside your chat. Think of it as your personal assistant with supercharged capabilities—whether for travel, wellness, education, or productivity. Join us as we unleash a new era of AI-driven utility in every conversation.
openai.com
Introducing Claude Sonnet 4.5
Claude Sonnet 4.5 sets a new standard in AI with unparalleled strength in building complex agents and exceptional proficiency at harnessing computers for advanced tasks. It’s designed to tackle intricate challenges, making it the most powerful coding model available today. This ultimate tool opens doors to innovative solutions and breakthroughs in technology.
anthropic.com
Build Your First (or Next) MCP Server with the TypeScript MCP Template
Learn how to set up your first or next MCP server using the TypeScript MCP Template. This guide simplifies the process, offering developers a solid foundation to get started quickly and efficiently. Perfect for those looking to streamline their server creation while leveraging TypeScript for robust and maintainable code.
nickyt.coPodcast
Talking AI
Talking AI is a podcast hosted by Matt Paige that explores the exciting world of artificial intelligence. Featuring engaging conversations with AI experts, product founders, and industry leaders, the show sheds light on how AI is transforming businesses and creating value. Perfect for both beginners and experts, each episode offers clear insights into how AI technology functions and its practical applications.
open.spotify.com