The AI in your browser
Drop an image and watch small open models read, caption and enhance it on your own device. Nothing gets uploaded.
/LABS · IDEAS IN PROGRESS
Open the
lab.
A home for our experiments, useful little tools and ideas we’re testing. Try something out. See what breaks. Help shape what comes next.
- ● OPEN · Ready to explore
- ○ INVITATION · Come test with us
The AI you didn’t know was in your browser.
Drop up to ten images. Your browser reads the metadata, flags what could give you away, captions the picture, finds the objects and faces, reads any text, then upscales, cuts out and fakes a lens blur. Small open models, running on your own device. The originals never get uploaded.
StackNext.js · Transformers.js · WebGPU · Florence-2 · MediaPipe
- [local]
- Reading IMG_4827.jpg
- [exif]
- 34 fields discovered
- [privacy]
- GPS coordinates detected
- [faces]
- 1 face found
- [model]
- 7 concepts detected
- [network]
- 0 bytes uploaded
EXPLORE THE EXPERIMENTS
- Try it
- Request
DQ Intel
Drilling intelligence for an energy client. Ada reads plans, scores her confidence and drafts the morning briefing.
- Request
Ops floor
The agent presence layer on this site as a drop-in script. Named cursors, selection frames, a live log.
- Request
Paper agent
The scripts that let an agent design in Paper. JSON-RPC client, write helpers, token recolour. How this site got built.
/ Access
Come and try.
Some experiments are open to explore. Others need an invitation while we test and refine them. Tell us what you’re curious about and we’ll take it from there.
- Keys
- Access details shared when your invitation is ready
- Data
- Labs run on synthetic or client-approved data only