Google Brings Agentic AI to Chrome 🌐
Chrome just became an AI agent. Google has unveiled a sweeping set of upgrades for its Chrome browser, effectively turning the world’s most popular web portal into an AI-powered operating system. The updates leverage Gemini 3 to not just browse the web, but to act on it.
Here is what’s new in your browser:
- Auto Browse: This agentic feature takes control of a tab to complete multi-step tasks—like booking tickets or researching flights—on its own. It handles clicks and navigation, pausing only for sensitive actions like payments.
- Gemini Sidebar: A new persistent sidebar lets you chat with Gemini alongside any webpage. You can ask it to compare products across multiple open tabs or pull data from your Gmail and Calendar without leaving the page.
- Nano Banana Integration: Google’s image model, Nano Banana, is now built directly into Chrome, allowing users to generate or edit assets instantly within the browser.
- Personal Intelligence: A forthcoming feature that will tailor search results and answers based on your personal data and browsing history.
Why it matters: Browser competition has heated up with entrants like OpenAI’s Atlas and Perplexity’s Comet. Google’s response is to muscle in with its massive ecosystem advantage. By baking agentic capabilities directly into Chrome, Google eliminates the need for users to switch to a niche “AI browser,” securing its dominance for the agentic era.
UrviumAI Take: Auto Browse is the feature to watch. You can test this on “admin” tasks. Next time you need to find a specific government form or compare flight prices across three dates, hand it off to Auto Browse. If it works, the days of “tab fatigue” are numbered.
China Gives Nod to Tech Giants to Buy Nvidia H200 Chips 🇨🇳
China reopens the door to Nvidia. In a major policy shift, the Chinese government has reportedly given the green light to its tech triumvirate, ByteDance, Alibaba, and Tencent, to purchase Nvidia’s H200 AI chips.
Here is the deal behind the scenes:
- The Approval: Beijing has authorized the import of over 400,000 H200 chips. Other firms are queuing up for similar permissions.
- The Visit: The regulatory nod coincided with Nvidia CEO Jensen Huang’s visit to China this week, highlighting the diplomatic weight of the semiconductor industry.
- The Balancing Act: While allowing these imports ensures Chinese tech giants stay competitive in the global AI race, Beijing is attaching strings. Reports suggest a “bundling” requirement where companies must also purchase a set ratio of domestic chips (likely Huawei’s) to support the local ecosystem.
- The Tech: The H200 is Nvidia’s second-most powerful chip, crucial for training frontier models. Without it, Chinese labs risk falling permanently behind U.S. rivals like OpenAI and Anthropic.
Why it matters: This is a pragmatic retreat from “total self-reliance.” Beijing realizes that starving its own tech giants of Nvidia hardware hurts China’s AI ambitions more than it helps its chip industry. It’s a calculated move to keep ByteDance and Alibaba in the game while buying time for Huawei to catch up.
UrviumAI Take: The “bundling” condition is the clever regulatory hack. Watch the adoption rates of Huawei’s Ascend chips alongside these H200s. If Chinese tech giants are forced to use domestic chips for inference while using Nvidia for training, it creates a unique hybrid architecture that could define China’s AI ecosystem for the next decade.
Two New AI Labs Raise Big to Rethink Learning 🚀
The “Contrarian AI” wave is here. Two new startups have emerged from stealth with massive war chests, betting that the current method of “feeding AI the whole internet” is a dead end. Flapping Airplanes and Core Automation are taking radically different paths to AGI.
Here is the scoop on the new challengers:
- Flapping Airplanes: This oddly named startup raised $180 million at a $1.5 billion valuation from GV and Sequoia. Advisors include legends like Andrej Karpathy and Jeff Dean. Their mission? Build AI that matches human intelligence without ingesting petabytes of data, aiming for 100,000x better data efficiency.
- Core Automation: Founded by ex-OpenAI researcher Jerry Tworek, this lab is seeking up to $1 billion. Tworek wants to solve “continual learning”—building a model called Ceres that learns from real-world experience in real-time (like a human does), rather than being frozen after training.
- The Vision: Both labs are betting against “scaling laws.” They believe smarter architectures, not just bigger clusters, are the key to unlocking true reasoning and eventually automating factories or terraforming planets.
Why it matters: The $200B+ labs (OpenAI, Google) are all doing the same thing: brute-force scaling. These startups represent the “rebel alliance.” Investors are pouring billions into them because if one of them cracks “data-efficient learning,” they could leapfrog the incumbents at a fraction of the cost.
UrviumAI Take: “Continual Learning” is the holy grail. Current AIs are “amnesic”, they don’t remember you from day to day unless it’s in the chat history. Core Automation’s Ceres model aims to fix this at the synaptic level. If they succeed, we will finally have AI that “grows up” with us.
Last AI News: OpenAI Launches Prism, Anthropic CEO’s Warning, & Microsoft Maia 200
Jigar Chaudhary is the Editor-in-Chief at UrviumAI, where he oversees coverage of artificial intelligence news, tools, and in-depth studies. With over 5 years of experience analyzing AI and robotics, he focuses on maintaining high editorial standards, accurate reporting, and clear explanations to help readers understand how AI is shaping the future.
