The Big Story

Google is letting anyone combine real satellite photos and 3D maps with AI image generation inside Google Earth on the web. You zoom into a spot, tap “create image,” and type a description like “what this street looked like in 1925” or “add a skate park here.” The system blends the real geography with your idea so the new scene still lines up with the actual buildings and terrain.

Think of it like taking a photo of your room and then asking an artist to paint something new on top while keeping the furniture exactly where it is. The AI doesn’t just make up a random picture; it respects the real layout so the result feels like it could actually exist.

This matters because maps used to be only for looking up directions or checking traffic. Now they become places where you can explore history, test ideas for your neighborhood, or create art projects for school. A student could show what their town might look like with more green space, or a family could see how their street changed over decades.

For you personally, it means a free way to turn geography homework or creative projects into something visual and shareable without needing special software. You can experiment right now by going to Google Earth on any web browser, finding a place you know, and typing a prompt about how you want it to look different. No account is required to start creating. The feature runs on Nano Banana 2 and works for every user starting today. Source: x.com

Explain Like I'm 14

You know how when you plan a big day out, you don’t just decide one thing at a time? You picture the whole sequence: first the bus, then lunch, then the movie, and you adjust if something runs late.

Now picture a robot trying to clean a room. It has to notice where the trash is, decide which hand to use, move its whole body without knocking things over, and check whether the floor is actually clean afterward.

The new Gemini Robotics ER 2 model acts like the robot’s “big-picture planner.” It watches the room with cameras, lists the steps needed, tells the movement system what to do next, and keeps checking progress. If a step fails, like a can rolling away, it notices and picks a new action instead of getting stuck.

This is different from older robots that could only follow one fixed set of moves. The planner lets the robot handle surprises the same way you would change your route if a sidewalk was closed.

Gemini Robotics 2 itself supplies the lower-level control for whole-body motion, dexterity, and even teamwork between multiple robots. The ER 2 layer sits above that control system and feeds it high-level instructions while tracking whether each instruction succeeded.

When the robot finishes one sub-task it does not stop and wait for new orders; it immediately decides the next logical step from its running list. That running list can include safety checks and recovery moves that were never written into the original plan.

The result is a robot that can start a job in a completely new room, notice when something unexpected happens, and still reach the final goal without a human rewriting the instructions.

Cool Stuff & Try This

Turn any real place into your own scene Google Earth’s new image tool lets you mix real maps with whatever you imagine. It’s perfect for school projects, art ideas, or just seeing how your street could look with different buildings or trees. Go to earth.google.com on your browser, zoom to any location, click “create image,” and type a short description. Try something like “add a giant treehouse” or “show this park in winter” and watch how the AI keeps the real roads and buildings in place. The tool is powered by Nano Banana 2 and is available to every user right now with no sign-up required.

A smarter way to follow recipes with AI A clear chart layout invented years ago lines up every ingredient in the exact order you need it and shows steps across a timeline so you never lose your place. It works like a visual checklist that removes the back-and-forth of regular recipes. You can ask any chatbot to turn a normal recipe into this chart format in seconds. Open ChatGPT or Gemini, paste a recipe, and type “convert this into a step-by-step chart with ingredients on the left and actions across the top.” The original layout was created in June 2004 by engineer Michael Chu and treats cooking like a Gantt chart so merged cells show when ingredients combine. Source: x.com

Quick Bits

Midjourney just released version 8.2 The popular image generator keeps getting better at turning text into detailed pictures. If you already use it for art or school projects, the update is live now.

OpenAI made some of its newest models much cheaper to use Lower prices mean tools like ChatGPT can handle longer or more frequent tasks without costing extra. Everyday users will notice the models feel more responsive in the free and paid versions. GPT-5.6 Luna dropped 80 percent in price and GPT-5.6 Terra dropped 20 percent, with a faster option added for GPT-5.6 Sol. Source: x.com

💬 Reply to this email — Patrick reads every one.

Share: X · LinkedIn · WhatsApp

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