CranleyGitHub

Tracks

Amazon: Best Practical AI Innovation

Ask in plain words and get a workspace you can act on. It works on your own home through the iPhone scan.

Judged creativity, usefulness, technical implementation and user experience, 25% each.

Angle. AI someone would use every day: a person with a real need about a building gets an answer they can act on, without learning a tool. And it works on their own home, not only on a dataset.

Criterion What serves it
Creativity An agent that builds the interface per need instead of a chat box; physics and geometry instead of text answers.
Usefulness The wheelchair, flood, fire, garage, pet and furniture tools answer questions people ask on forums every day. The iPhone scan makes it work on any building.
Technical About 600 server tests and 90 web tests; every tool run on a 15-place demo set with a matrix of results (see Proof); FastAPI, deepagents, two cloud databases, our own 3D renderer and splat training.
User experience The App look: one thing first on every screen, the rest folded; plain-word reliability chips; every panel resizable; guided first run; both themes; phone companion with live sync.

Adaptive intelligence (a named focus area): the agent’s memory per building and per person, “what I learned about you” with undo per item, past investigations searchable by words and meaning.

Scripted flow: web/src/skins/app/demo/flows/amazon.json, “A phone scan becomes a PDF report”.

Demo path. Phone: scan a room with the iPhone app (demo cafe in the simulator), name the business, pick Accessibility, get the PDF in seconds. Desktop: the same workspace updates live. Then the wheelchair search above.

The demo

The Cranley iPhone app showing a scanned plan and the tools run on it (1 of 3)The Cranley iPhone app showing a scanned plan and the tools run on it (2 of 3)The Cranley iPhone app showing a scanned plan and the tools run on it (3 of 3)
The iPhone app in the simulator, on the made-up sample home: the plan from a RoomPlan export, the tools run on it, and what the scan cannot tell.

In the docs