The Big Story

OpenAI announced it will start adding special hidden markers called text watermarks to some of the writing that ChatGPT and its other tools produce. These markers are invisible to humans but can be detected by special software. Think of it like a secret stamp that only certain readers can see — the stamp doesn't change how the words look or feel when you read them. The goal is to help people know whether a piece of text probably came from an OpenAI model, which is becoming important as more students, writers, and everyday users create content with AI. This change is happening first in Europe to follow new rules about AI transparency, and it will roll out to ChatGPT and Codex text over the coming weeks. For you, it means that when you're using AI to help with homework, stories, or social media posts, there could be a way for others to check if AI helped. That might feel a little strange at first, but it also makes the whole system more honest and could reduce worries about fake or misleading content. Right now the detector tool is only open to approved researchers so they can test and improve it, but the idea is to give people more choice and clarity about what AI can and can't do. You can read OpenAI's own explanation on their site to see the exact details. Source: openai.com

Explain Like I'm 14

You know how when you're texting with friends, your phone sometimes predicts the next word based on patterns it has seen in millions of other messages? Now imagine that instead of just guessing one word, the system adds a very faint, repeating pattern across the whole message — like lightly sprinkling invisible glitter that only a special scanner can pick up. The pattern isn't made of extra letters or symbols you can see; it's a tiny statistical nudge in which words get chosen at certain moments during generation. A detector tool later looks for that nudge the same way you might notice a secret rhythm in a song if you had the right headphones. Because the pattern lives in the choices the model makes while writing, things like rewriting the text or translating it can wash the pattern away, which is why OpenAI is being careful about how far they trust the system right now. The result is a quiet way to answer one simple question: was this text likely made by one of our models? It doesn't reveal who wrote it or own it — it just gives a clue about its origin. Source: x.com

Cool Stuff & Try This

TikTok's new shopping helper that chats with you TikTok just added a conversational AI agent inside its app that can help you find products, answer questions about them, and even let you check out with one tap. It works like having a helpful friend who knows the entire store catalog and can suggest things based on what you describe. This is exciting because shopping on social media is already something lots of teens do, and now the AI can make suggestions feel more natural and personal. Anyone with the TikTok app can try it right away. Open the app, go to the shopping section, and start a conversation by typing something like "show me cool sneakers under $50" to see how the assistant responds and guides you. Source: techcrunch.com

Instinct lets friends team up with an AI agent in group chats Instinct is rolling out group chats where multiple people can work together with the same AI agent on tasks like planning a trip, organizing a carpool, or setting up an event. Each person keeps their own private account, and the agent only shares information or takes action after everyone agrees. It's a fun way to see how AI agents can join real conversations instead of working alone. You can try it by visiting the Instinct website or app and starting a group chat with friends who don't even need their own accounts to participate. Source: techcrunch.com

Quick Bits

HackerRank's AI that practices job interviews with you HackerRank built an AI interviewer that has already run over 500,000 practice sessions and is being tested by companies like Snowflake and Capgemini. It gives students and job seekers a low-pressure way to rehearse answers and get feedback before real interviews. Source: techcrunch.com

💬 Reply to this email — Patrick reads every one.

Share: X · LinkedIn · WhatsApp

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