You’ve seen what 𝗚𝗣𝗧-𝟲 𝗔𝘀𝘁𝗿𝗮 can do for images, videos, 2D and 3D worlds. But what about a dimension beyond the flat screen? We wanted to put it to the test in spatial XR. But how? We directed an agent to build a Jarvis-style Earth interface that visualizes millions of real-time data points on @PICOXR’s new device, where you can interact with it. All real in-headset footage: ✈️ Live air traffic, earthquakes, satellites and the ISS, all updating in real time 🖐️ A globe you turn with your hands 🏙️ Tap any place and it becomes a 3D city on your table 📊 4 data panels open side by side in one app Why this matters: building for spatial is a different kind of problem from building for a flat screen. The model has to reason about depth, scale, hand input and layout in 3D space. Astra wrote the app from scratch as a native 𝗞𝗼𝘁𝗹𝗶𝗻 app using the PICO CLI, and the 3D cities stream in from @Google Photorealistic 3D Tiles via @CesiumJS A year ago, this would have taken a team months. Now one person can direct an agent and then walk around the result. Inspired by @bilawalsidhu’s open-source 𝗚𝗼𝗱’𝘀 𝗘𝘆𝗲 𝗩𝗶𝗲𝘄 and vibe-coded by @nigelhartmanxr How far can coding agents really go in 3D and spatial? We’re testing that live in our 𝗳𝗿𝗲𝗲 𝗼𝗽𝗲𝗻 𝗹𝗲𝗰𝘁𝘂𝗿𝗲 𝘁𝗵𝗶𝘀 𝗪𝗲𝗱𝗻𝗲𝘀𝗱𝗮𝘆. Bring your own app idea, and let’s talk about what frontier models can (and can’t) build. Link below👇 #AI #AgenticAI #XR #SpatialComputing #PICO

No way! 🤯 @lucas_martinic is pitching his vibe-coded app @Polyfork3d to @PalmerLuckey, @xprize and @Google leaders this Friday. Out of 26,000+ builders, he made the TOP 5, and he's closing in on a share of the $2M prize pool ✨