The Big Story

Google released a new version of its Gemini Omni 1.1 Flash model that makes creating AI videos easier and less expensive. Instead of only looking at the last second of video to continue a scene, the model can now look back at up to ten seconds of footage. This helps keep characters, backgrounds, and actions consistent when you want to extend a clip.

Think of it like editing a short movie on your phone: you pick a clip, and the AI helps you smoothly add the next part without the picture suddenly changing style or the people looking different. Scenes can now be extended in 10-second chunks up to 40 seconds total. There's also a new 360p draft mode that runs up to 60 percent faster and costs about a third as much as before.

The change matters because earlier tools often produced jumps in style or appearance after just a few seconds, making longer clips feel broken. With the longer look-back window the model sees more of the original lighting, movement, and objects, so the added footage stays matched. For students making projects or creators experimenting with short stories, this lowers the barrier to trying video ideas without waiting forever or paying a lot.

If you've ever wanted to turn a quick prompt into a longer scene for a school video or TikTok-style post, these changes make it more practical. The updates are already rolling out in Google AI Studio and the Gemini app for subscribers. You can head to Google AI Studio on your laptop or phone, sign in with a Google account, and start testing the new video tools with your own short prompts right away. The cheaper draft mode lets you preview ideas quickly before committing more time or credits to a finished version. Source: the-decoder.com

Explain Like I'm 14

You know how when you’re watching a short video on your phone and you want to keep the story going, you might imagine what happens in the next few seconds based on what you just saw? AI video tools do something similar, but they do it by looking at a chunk of the video that already exists.

The model studies the last several seconds of movement, lighting, and objects so the new part matches what came before. It then predicts the next frames the same way a really good autocomplete guesses the next words in a sentence you’re typing. When the model can look farther back — like ten seconds instead of one — it has more context, so characters stay the same size and the background doesn’t suddenly shift.

That extra look-back window is what lets scenes feel like one continuous take instead of random jumps. The cheaper draft mode simply uses fewer details while it builds the clip, so you get a fast preview before spending more time or credits on the polished version. Because the model now has a longer memory of the original clip, the added seconds are more likely to keep the same colors, camera angle, and character appearance. This is why the 10-second window and the 360p draft option together make longer videos feel realistic without extra cost.

Cool Stuff & Try This

Gemini Notebook now chats with your books Google’s note-taking app called Gemini Notebook just added a feature that lets you pull in books you’ve bought from Google Play Books. You can ask questions about the content, request study plans, or even generate simple infographics and short audio summaries based on the chapters.

This turns the app into a personal tutor that already knows the exact book you’re reading for class or for fun. It’s especially useful if you’re studying history, science, or literature and want quick explanations without flipping back through pages. The new “Expert Intelligence” feature brings titles from Google Play Books directly into the notebook so the AI can reference specific passages.

Go to the Gemini app or Gemini Notebook on your phone or computer, connect your Google Play Books account, pick a title you own, and try asking something like “Explain the main idea of chapter 3 in simple terms.” You can also ask it to create a study plan for an upcoming test or turn a chapter into a short audio summary you can listen to while walking between classes. Source: theverge.com

Quick Bits

Realtime AI video that finishes before you finish watching A new tool called H3 Max can now generate decent-quality video clips in less time than it takes to watch them, including automatic prompt improvements. Early tests show it crossing into truly fast, usable territory for quick experiments. The speed comes from generating the video in realtime from the moment the generate button is pressed, so the whole clip appears before the original prompt has even finished playing back. Source: x.com

Big tech companies team up on AI safety OpenAI, Google, Anthropic, and over 100 other groups published an open letter asking for stronger shared defenses against AI-powered cyberattacks on things like hospitals and power systems. The idea is to give defenders better tools before problems get worse. The letter warns that AI advances create a limited window to strengthen cyber defenses and calls for collective action across companies including AWS, Microsoft, and Oracle. Source: techcrunch.com

```claims []

💬 Reply to this email — Patrick reads every one.

Share: X · LinkedIn · WhatsApp

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