been having fun making quick webgl sketches and toys with Gemini 4 Argon. I screencapped them and it edited together this reel. then I described the soundtrack I wanted and it procedurally composed it in raw C++ math... just to flex on me I think

Meet Gemma Translator: a completely offline, DIY translation device powered by Gemma 4 and built with Antigravity. 🧵

Code Pictionary ✏️🧠 AI draws something. You try to guess what it is. Made with Gemini 2.5 Flash as a fun way to see how AI illustrates different concepts. Link in 🧵

🌡️✅ been building/using this tool to test out new models and get a quick, visual sense of their capabilities 👇

Veodyssey 🦡🔱🔊 A prototype experimenting with new ways to learn about The Odyssey. Made w/ @dcmotz using Veo 3 + Gemini 2.5. Try it here: veodyssey-v0.appspot.com Details + prompts in 🧵

We're using Gemini to help write prompts. ✏️ Really excited about interfaces like this quick one @dcmotz made with Gemini API. It lets us iterate quickly, finding ideas we like, and feed them back in. Rapid human-in-the-loop co-thinking. We can share this tool if people are interested in using/forking it. We've gotten best results when we let Gemini give directions but doesn't prescribe verbatim dialogue. This gives Veo more creative freedom and the outputs tend to be better and more coherent.

New Gemini 2.5 Pro preview dropped today, had it make this quick sketch where it draws its own logo with a Fourier series 😎✏️📐 g.co/gemini/share/172aa74e75…

Want to play with Gemini + multimodality? ✏️🧠 Here's a new "Multimodal Canvas" demo by @dcmotz built w/ Gemini 1.5 Flash that lets you quickly test prompts using drawing, camera, images, emojis: labs.google/multimodal-canva… One of my favorite things to try is pictionary: