In partnership with

🧩 Anthropic’s New Chrome Extension Pushes Limits and Raises Questions 

Anthropic has begun piloting Claude for Chrome, an extension that allows its AI assistant to work directly in your browser. For some users, it’s a breakthrough step in making Claude part of everyday browsing. For others, it’s raising red flags about whether AI agents can be tricked by malicious sites.

🔧 What the Extension Does

The extension brings Claude into Chrome’s sidebar, letting you summarize pages, draft emails, or answer questions while you browse. Early instructions from Anthropic’s support guide explain how to install and log in. Current access is limited to select U.S. users, though wider rollout is planned in the future. Users can toggle Claude on and off per site, and permissions are designed to minimize background monitoring.

🚀 Why It’s a Major Step

Embedding Claude directly into Chrome could make it a true everyday copilot, similar to Microsoft’s Copilot in Edge. Instead of pasting links into chat, users can highlight content and get contextual answers instantly. Many are calling it a significant test of how tightly AI assistants can integrate into mainstream browsers. It also signals Anthropic’s push to compete not just on model quality, but on accessibility and seamless workflow.

⚠️ The Hijacking Risk

Security researchers warn that AI agents in browsers are vulnerable to hidden instructions embedded in websites. These invisible prompts in HTML or metadata could mislead AI tools into leaking sensitive data or performing unwanted actions. While Anthropic stresses Claude’s Constitutional AI training as a safeguard, experts note attackers are constantly finding new prompt-injection techniques. The risk is heightened when an AI tool has page-level context and can trigger follow-up actions on a user’s behalf.

🔮 What Comes Next

For now, Claude’s Chrome extension is limited to 1,000 Max Plan users. If you’re interested, here's the link to sign up for the waitlist. If this rollout proves successful, it could mark a shift where every major AI player moves from chat windows into browsers.

Google is rumored to be exploring similar integration for Gemini, and OpenAI has long been linked to a browser project. The real test will be whether safety frameworks keep pace with creative misuse—and whether users feel comfortable trusting an AI that sits inside their daily browsing.

The Key to a $1.3 Trillion Opportunity

A new trend in real estate is making the most expensive properties obtainable. It’s called co-ownership, and it’s revolutionizing the $1.3T vacation home market.

The company leading the trend? Pacaso. Created by the founder behind a $120M prior exit, Pacaso turns underutilized luxury properties into fully-managed assets and makes them accessible to the broadest possible market.

The result? More than $1B in transactions and service fees, 2,000+ happy homeowners, and over $110m in gross profit to date for Pacaso.

With rapid international growth and 41% gross profit growth last year alone, Pacaso is hitting their stride. They even recently reserved the Nasdaq ticker PCSO.

The same VCs that backed Uber, eBay, and Venmo also backed Pacaso. Join them as a Pacaso shareholder before the opportunity ends September 18.

Invest While You Still Can

_{Paid advertisement for Pacaso’s Regulation A offering. Read the offering circular at invest.pacaso.com. Reserving a ticker symbol is not a guarantee that the company will go public. Listing on the NASDAQ is subject to approvals.}

🎬 Google Vids’ New Upgrades Are Impressive

Google has rolled out major upgrades to Google Vids, making it free for everyone while expanding its collaborative video tool with AI avatars and image-to-video generation for paid users. The goal: make professional-quality content creation faster and more accessible.

🧑‍🎤 AI Avatars Enter the Scene

Users can now create talking avatars that read scripts with natural lip-syncing and gestures. The avatars draw on Gemini’s language and multimodal models, enabling quick corporate explainers, training videos, or personalized messages.

🖼️ Image-To-Video Capability

The other standout feature: turning static images into moving video clips. Early demos show smooth transitions like zooming into a still photo or animating backgrounds. The feature is aimed at marketers and educators who want more dynamic visuals without expensive production.

🛠️ How to Use It

Google has updated its support documentation with step-by-step instructions. Previously, Google Vids was only available to those with access to a Gemini-enabled Workspace. However, with the latest update, anyone can use it.

🌍 The Bigger Picture

The updates reinforce Google’s strategy of embedding Gemini across its productivity apps. With Microsoft pushing Copilot Studio and startups offering video-first platforms, Google Vids now looks less like a side experiment and more like a central pillar of its enterprise suite. 

This Week’s Scoop 🍦

📊 Stanford study shows entry-level workers hit hardest by AI 

🥷 Criminals carrying out sophisticated cyber-attacks using Claude 

💡 Nvidia earnings highlight boom and possible bust for AI chips 

🧑‍💻 xAI introduces Grok Code Fast 1 

🌋 Tokyo using AI video to help citizens prepare for Mt Fuji eruption 

📽️ Taiwan's Central News Agency (CNA) showcases AI news agent

🎯 Weekly Challenge: Prompt Swap With a Friend

Here’s what to do:

✍️ Step 1: Write your task. Each of you chooses one task that actually matters to you — it could be “draft a travel itinerary,” “summarize a long report,” or “create a recipe plan.” The key is to pick something real, not hypothetical, so the output feels useful.

🔄 Step 2: Swap prompts. Exchange your task with a partner and ask an AI model (ChatGPT, Claude, Gemini, etc.) to complete their task, not your own. By stepping into someone else’s shoes, you’ll see how much prompt style changes the outcome.

🔍 Step 3: Compare results. Look at what the AI delivered. Did it interpret your friend’s wording differently than they intended? Did your phrasing lead to answers that were sharper, flatter, or just plain different? Notice where the model stumbled versus where it surprised you.

💬 Step 4: Debrief together. Talk through what worked and what didn’t. Refine the prompts until you both agree on the strongest version, then rerun it to see if the AI converges on something better. This back-and-forth will reveal how even small word changes steer results.

🌟 Why it matters: Prompt writing is part art, part science. Swapping prompts exposes blind spots in your own style and helps you understand how others frame requests. It’s also a fun collaboration test — two brains plus one model often unlock better, more creative answers than either alone. The lesson: good prompts aren’t written, they’re co-created.

Want to sponsor Jumble?

Click below ⬇️

From AI agents browsing side-by-side with you to the explosion of AI video generators and tools, AI never sleeps. See you next time! 🚀

Stay informed, stay curious, and stay ahead with Jumble!

Zoe from Jumble