Last weekend I had coffee at the wonderful Uptown Espresso with a friend from a nearby Amazon office. I write "a" nearby office because he's working on a multi-year project so secret that he can't even tell people which building he's in.

Being a good corporate citizen, he wouldn't tell me a damn thing so I made up a story about what I hope he's making: On demand fabrication and assembly.

Basically, Amazon adds 3D printers to their existing picking system and workers walk the floor like they do now, but instead of picking products and throwing them into a box, they're picking parts and 3D renders which they first assemble into one-off products and then they throw them into a box.

The huge flock of product designers we're growing in universities could use web tools to create new product assemblies using both custom 3D printed parts and prefab modules like servos, controllers, and drive trains from "old school" manufacturing groups.

Specifically, a DIY drone hacker could use a web based CAD tool to design a 3D printable drone body around sensors and actuators which they pick from the Amazon Marketplace. Once the design works, the hacker could offer it in the Marketplace as a new product. It's like CafePress but instead of mugs and t-shirts with your drunken cat on them it's small run submarines and seriously niche jewelry.

This is exactly the sort of play which would leverage Amazon's existing plant and code but takes some serious scale so your average tech startup isn't going to be able to attract enough funding. Companies like Shapeways have blazed a trail in 3D printing so Amazon has the second mover advantage for everything except the assembly tech and Seeed Studio has tried most of the naive solutions for that.

The software engineering effort is pretty massive. On the user facing side, Amazon would need to create a large list of products like tutorials and training apps, WebGL based CAD, and a marketplace for assemblies. Internally they'd need to build some serious tech like a damn good print and assembly simulator to prevent impossible designs, assembly instructions for workers, and moderation to prevent people from shipping drones and guns (inside the US, anyway).

When I'm tempted to leave the startup world and join a big co, this is the sort of project to which I am attracted.