I built SH-01 for the Meta Global AI Developer Hackathon.
It's an AI assistant that doesn't just tell you what to do. It shows you where to click, on your own screen.
I used it to build a small game in Godot, an app I had never used.
Open source: https://t.co/iFN693oYSc
Try it: https://t.co/x22HmR9ymw
How it works:
1. It takes a screenshot of your screen
2. Meta's Muse Spark finds the right button and builds a plan
3. Press Show me, and the robot points at it
4. Press Next, and it checks your screen to confirm the step is done
What is SH-01?
It's a macOS AI assistant that guides you inside any app. Tell it your goal, and instead of a wall of text, a small pixel robot flies to the exact button you need and points at it. Step by step.
I built SH-01, an AI assistant that doesn't just tell you how an app works. It shows you where to click, right on your screen.
I used it to build a small game in Godot, an app I'd never used.
Try it: https://t.co/x22HmR9ymw
Code: https://t.co/iFN693oYSc
Ideas for what to add next? I'd love to hear them.
My bet for the hackathon was simple: the next interface isn't more text it's an AI that looks at your screen and shows you the way, step by step.
SH-01 does it live with Muse Spark vision grounding. Nothing hardcoded.
Demo: https://t.co/qeBPflq8ip
Built solo in 10 days for the Meta Global AI Developer Hackathon. Open source, Apache 2.0.
π¬ Demo: https://t.co/qeBPflq8ip
π Site: https://t.co/NVLCsL90pu
π» Code: https://t.co/smh2BaFlZU
Built solo in 10 days for the Meta Global AI Developer Hackathon. Open source (Apache 2.0).
π Site: https://t.co/NVLCsL90pu
π» Code: https://t.co/smh2BaFlZU
All you need: a Mac + a Meta Model API key.
π Launching DeepSeek-V3.2 & DeepSeek-V3.2-Speciale β Reasoning-first models built for agents!
πΉ DeepSeek-V3.2: Official successor to V3.2-Exp. Now live on App, Web & API.
πΉ DeepSeek-V3.2-Speciale: Pushing the boundaries of reasoning capabilities. API-only for now.
π Tech report: https://t.co/7EyydyNuG0
1/n