• Fixed a bug with the wireless camera. • Fixed an issue that could cause block programs to stop running. • Fixed an issue with signing in using a Google account.

What's New We’ve fixed various bugs to make camera-based model inference more stable and faster. New blocks are here! You can now use dedicated blocks to directly control the running state of programs on the home screen. Blocks also get smarter: customizable conditions let makers build more complex logic with ease. Pose Classification just got a nice accuracy boost during model training — give it a try!

External Camera Support Connect external Wi-Fi cameras (including ESP32-based devices) and use them directly as the AI vision source — stream live feeds into the inference pipeline alongside or instead of your built-in camera. AI Agent Mode Design custom AI workflows with a drag-and-drop node editor. Chain together camera snapshots, LLM reasoning, memory, voice output, and Bluetooth device commands in any configuration. Build agents that see, think, remember, and act — entirely on-device. Gesture Classification Train your own hand gesture models without writing any code. Collect samples, train instantly on-device using KNN, and put your custom gestures to work in programs and AI workflows right away.