Hey Product Hunt! 👋

I'm Joshua, the creator of Sightline CRDT.

As computer vision models have grown exponentially, the way teams audit and verify training datasets hasn't kept up. Most ML engineers are stuck with clunky single-player desktop tools from a decade ago or paying tens of thousands of dollars to slow labeling agencies.

I built Sightline CRDT to give vision teams a "Figma for Computer Vision Datasets":

✨ Zero-latency peer-to-peer multiplayer using Yjs CRDTs over WebRTC

🎯 Sub-pixel bounding box manipulation & uncertainty triage

🎙️ Multimodal hands-free voice patching for 3x faster annotation

⚡ 100% in-browser, privacy-first, zero server lock-in, with instant COCO/YOLO export

Try the live sandbox: https://sightline-crdt.netlify.app

I would love to hear your thoughts, feedback, and what features you’d like to see next! What’s your biggest bottleneck when auditing training data?

Thank you for checking it out! 🚀