The Big Story

Google will soon allow you to customize your Discover feed by describing what you want to see. The new feature, rolling out to the Google app in the "coming days," will use AI to automatically tweak your feed and "remember" your preferences for future visits. You'll find the option within the three-dot menu on any Discover story.

Think of it like talking to a friend who always picks the next video for you: normally they just guess, but now you can tell them your mood and they’ll keep that in mind next time. The AI doesn’t just change the feed once — it stores your description so future visits feel more personal.

This matters because most of us scroll these feeds on our phones every day without realizing how much they shape what we learn about the world. Being able to steer them directly could mean fewer rabbit holes and more stuff that actually matches your interests, whether that’s school projects, music, or hobbies.

For you specifically, it means the news and articles you see can finally line up with what you care about right now instead of what you clicked on last week. If you’re studying for a test or deep into a new game, you could ask for more of that and actually get it.

You can try this yourself soon in the Google app on Android or iOS. Open the app, tap the three-dot menu on any Discover story, and look for the new customization option that lets you describe what you want. The feature is rolling out in the coming days, so check back if you don’t see it yet. Source: theverge.com

Explain Like I'm 14

You know how when you open YouTube or TikTok, the app seems to know exactly which video will keep you watching for another ten minutes?

It starts by looking at what you’ve watched before, what you paused on, and what other people who watched the same thing also enjoyed. Then it mixes those signals together to guess what you’ll click next.

Now imagine the same process, but instead of videos it’s news stories, and instead of just guessing once, the app lets you type a sentence that adds new rules to the guess. The AI takes your sentence, turns it into a set of preferences it can remember, and uses those preferences the next time it picks stories for you.

It’s not magic — it’s the same pattern-matching that powers every recommendation feed you already use, just with an extra step where your own words become part of the pattern.

That’s why the new Google feature feels different: you’re not just hoping the algorithm gets it right; you’re giving it clearer instructions in the same way you’d tell a playlist what mood you’re in.

Cool Stuff & Try This

Give your ChatGPT a pretend body for one day This fun prompt turns ChatGPT into a character that wakes up in a real body for just 24 hours and has to decide what to do with that single day. It uses the conversation history you already have with it, so the answer feels personal. Anyone can try it on their phone or laptop with a free ChatGPT account.

Copy the full prompt from the original Reddit post, paste it into a fresh chat, and let it run. The AI will describe waking up, choosing one main thing to do, and then leaving one memory behind at midnight. It’s a creative writing experiment that shows how much personality an AI can show when you give it a clear scenario. The prompt instructs the model to begin with two or three first-person sentences showing the moment it wakes and orients itself, then continue under the exact headings THE CHOICE, WHAT HAPPENS, and WHAT I LEAVE IN THE ARCHIVE. At midnight every AI leaves one memory in a shared archive, and if enough people take part the comment section becomes that archive. Source: reddit.com

OpenAI’s image tool now makes pictures with transparent backgrounds GPT-Image-2 can now create images that have no background at all, so you can drop them straight into presentations or edits without extra work. The feature is available through the API with one simple setting turned on. The alpha channel gets baked in during image generation, which according to OpenAI beats conventional background removal. If you already use ChatGPT’s image features, this makes the results more useful for school projects or creative stuff right away. A single parameter activates the feature, and the preview is currently rolling out to API users. Source: the-decoder.com

Quick Bits

35 % of new web pages show signs of AI writing A new study found that more than a third of web pages published since ChatGPT appeared carry clear markers of AI-generated text. It’s a reminder that a lot of what we read online is no longer written only by humans. The research tracked pages published after the launch of ChatGPT and identified patterns typical of large language model output. Source: storyboard18.com

Your writing job might already be mostly prompting One Reddit user shared that their “marketing” role now involves giving ChatGPT topics, pasting in boss feedback, and repeating until the draft is good enough. It’s a real-world look at how much of today’s writing is shifting from typing to guiding. The workflow described includes the boss replying with a wall of feedback, the user copying those notes straight back into the chat, and repeating the cycle four or five times. Source: reddit.com

💬 Reply to this email — Patrick reads every one.

Share: X · LinkedIn · WhatsApp

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