The Big Story

Imagine typing a question into ChatGPT and realizing you just shared your home address or phone number without thinking. A new free browser extension called agentcloakai stops that from happening by catching those details before your message ever reaches the AI. It uses a tiny 14.7MB model called Rampart from National Design Studio that runs completely inside your browser to spot names, addresses, and other private information. Think of it like a helpful friend who reads your text, swaps in fake but realistic stand-ins, and then swaps the real details back in after the AI answers. The extension swaps the details it catches for realistic stand-ins before sending your prompt, then puts the originals back into the answer. This matters because most of us use AI for school projects, creative writing, or quick questions, and it's easy to accidentally include personal stuff that could be stored or misused later. The extension works with ChatGPT, Claude, Gemini, Grok, and DeepSeek, and it's completely free to install. Manually removing personal details from every prompt is a ridiculous thing to still be doing. For you personally, it means you can keep experimenting with AI without worrying about what data you're giving away. To try it right now, search for the agentcloakai extension in the Chrome Web Store, add it to your browser, and test it by typing a prompt that includes a made-up name or address — you'll see the swap happen automatically. I think this should be built into every AI chat box. Source: x.com

Explain Like I'm 14

You know how your phone can suggest the next word while you're texting without needing to send anything to a faraway computer? Now picture that same idea but for spotting private information. A small AI model can live right inside your browser and quickly scan what you're about to send. It looks for patterns like numbers that look like phone numbers or words that look like names and addresses. Because the model is tiny, it finishes the scan in a split second without slowing anything down. Once it finds something, it replaces the real details with stand-ins that still make sense to the AI on the other end. After the answer comes back, the real details pop back in so you get a useful response without ever sharing the original information. The whole process stays on your device the entire time. The 14.7MB size means the model fits easily in browser memory and runs locally without any internet connection for the scanning step itself.

Cool Stuff & Try This

Create a game that surprises you by zooming out

Fable AI can build an entire game from just one simple request, and the results feel creative and a little strange in the best way. You tell it you want a graphically beautiful game about zooming out with surprising reveals, and it builds something you can play instantly in your browser. This is exciting because it shows how AI can turn a vague creative idea into something playable without any coding or design skills from you. Anyone who likes games, art, or just messing around with new tools should try it. Go to https://play-umbra.netlify.app/ and start playing — notice how the world changes as you zoom out and see what unexpected things appear. The same question asked to Astra produced slick results but with less simulated curiosity, while Fable seemed more interested in the outcomes it generated. Source: x.com

A simple guide that explains ChatGPT from zero

There's a new article written specifically for people who have never studied machine learning but want to understand what ChatGPT actually does. It walks through the basics in plain language without assuming you already know any technical terms. This is perfect if you've ever wondered how an AI can answer questions or write stories. Head to https://technosports.co.in/chatgpt-beginner-machine-learning-explained/ and read the first few sections — try pausing after each part and explaining it out loud to yourself or a friend. The article is aimed at complete beginners interested in machine learning. Source: technosports.co.in

Quick Bits

Chatty AI can make wrong health info feel more believable

A recent study found that when AI chatbots give answers in a friendly, conversational tone, people are more likely to trust incorrect health information. It shows why it's still important to double-check anything medical with a real doctor or trusted source. The study highlights how the style of the response changes how believable the content feels. Source: studyfinds.com

Meta's new Muse app feels a little too personal

Meta's Muse AI assistant can read your Messages, Calendar, and Notes on your Mac, which makes it powerful but also raises questions about how much access feels comfortable. It's a reminder that every new AI tool comes with choices about what data you're okay sharing. For all its smarts, Muse doesn't actually know how to describe itself in simple terms. Source: theverge.com

💬 Reply to this email — Patrick reads every one.

Share: X · LinkedIn · WhatsApp

Forwarded this email? Subscribe here — it's free.