Global AI Weekly
Issue number: 123 | Tuesday, November 11, 2025
Highlights
Brace Yourself: ChatGPT Ads Are Rolling In
OpenAI CEO Sam Altman discusses the company's approach to ads and commerce with ChatGPT, contrasting it with Google's ad model. He highlights how Google's model benefits from ineffective search results and emphasizes that ChatGPT's ads will take a different direction. OpenAI aims to create a system that prioritizes user experience and efficiency without relying on the shortcomings of traditional search engines.
searchengineland.com
Tech stocks suffer worst week since April after $800bn AI sell-off
Tech stocks have faced their toughest week since April, with the Nasdaq dropping 3% over five days amid concerns over high valuations in the sector. The AI boom, which had driven a surge in market value, saw a sharp $800 billion sell-off as investors reassessed growth potential. This downturn comes as fears grow over whether current valuations can be sustained.
ft.com
Breaking the Million-Token Barrier
Azure's ND GB300 v6 Virtual Machines, equipped with NVIDIA GB300 NVL72 rack-scale systems, deliver impressive performance, surpassing 1,100,000 tokens per second on Llama2 70B. This milestone showcases the incredible capabilities of these systems in handling complex AI workloads efficiently. The achievement highlights the potential for accelerating innovation in AI and machine learning applications.
techcommunity.microsoft.comResearch
A Cross-Regional Study of AI Risk Management Frameworks in the EU, U.S., UK, and China
This paper examines AI risk management frameworks across the EU, U.S., UK, and China, exploring how each region balances the need for innovation with regulatory oversight. It provides a comparative analysis of policies and approaches, highlighting similarities and differences in addressing AI-related risks. The study aims to offer insights into global strategies for managing AI while fostering technological growth responsibly.
arxiv.org
Diffusion Language Models are Super Data Learners
This paper page discusses how diffusion language models excel as super data learners. It explores their ability to efficiently process and understand data, highlighting their potential in advancing machine learning and natural language processing tasks. The discussion sheds light on their performance and implications in various applications.
huggingface.coVideo
Real-Time Continual Learning for AI Agents - Jarrod Barnes
Arc is a real-time continual-learning framework designed to help AI agents adapt and improve based on experience. This talk highlights key developer techniques such as online prompt optimization and reward modeling for production environments, alongside a demo of the open-source SRE Agent learning to resolve a complex Kubernetes incident. Led by Jarrod Barnes, Co-Founder of Arc Intelligence, the presentation draws on his extensive experience in developer ecosystems, infrastructure scaling, and academic research on skill acquisition.
youtube.com
How Claude is transforming financial services
Anthropic’s Alexander Bricken and Nick Lin discuss Claude's capabilities tailored for financial services, emphasizing its ability to handle finance-specific tasks like real-time market analysis and creating Excel spreadsheets. They share examples from companies such as NBIM and BCI, highlighting how Claude streamlines workflows by conducting research, performing analysis, and producing investor-ready reports without the need to switch tools. They also outline Claude's adaptability, ongoing improvements, and future plans for deeper integrations and expanded applications within the financial sector.
youtube.comArticles
Inside Atlas: How OpenAI’s OWL Redefines the Browser for AI Copiloting
In their deep dive on the architecture of ChatGPT Atlas, OpenAI reveals how the new layer dubbed OWL (OpenAI’s Web Layer) decouples the Chromium engine from the main application process—enabling instant startup, smoother performance across hundreds of tabs, and richer agent-based browsing. By blending native UI frameworks with Chromium’s robust web platform, Atlas is positioned not just as a browser but as an AI-powered co-pilot navigating the web.
openai.com
The State of AI 2025: From Widespread Use to Scarce Transformation
The latest global survey by McKinsey & Company reveals that although 88% of organizations report regular use of AI, only one-third have reached enterprise-scale deployment. Interest in AI agents is on the rise, with 62% of respondents experimenting, but deeper value remains elusive: just 39% see a measurable EBIT impact. High-performing organisations distinguish themselves by redesigning workflows, aligning senior leadership, and pursuing innovation, showing that scaling AI isn’t just about tech, but about transformation.
mckinsey.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
langchain-ai/deepagentsjs: Deep Agents in JS
Deep Agents in JS is a project hosted on GitHub under langchain-ai, focusing on the development and implementation of advanced agents using JavaScript. The repository allows developers to contribute and collaborate, fostering innovation and growth in the field. It serves as a central hub for creating, sharing, and improving intelligent agent frameworks in JavaScript.
github.com
How to use GitHub Copilot from the command line
Learn how to use GitHub Copilot directly from your terminal with this guide to GitHub Copilot CLI. This guide includes a starter kit filled with effective prompts to help you tackle various tasks and streamline your workflows. Perfect for those looking to enhance their coding experience right from the command line.
github.blogPodcast
AI in Real Life: Real Talk with Real Experts
AI in Real Life: Real Talk with Real Experts is a student-led podcast from the AI Institute for Societal Decision Making at Carnegie Mellon University. The show brings together undergraduate and high school students to interview AI experts on topics that matter to the public. It offers accessible discussions on artificial intelligence, bridging the gap between experts and everyday audiences.
open.spotify.com