I love this idea. Two years ago, when I had just moved to Silicon Valley, I came up with a concept called OpenAnno for a hackathon about Twitter's new annotations API feature. Lucky for me (I didn't realize this then, I didn't know a lot about them), two of the judges were Ron Conway and Paul Graham. I naively presented the project and later ended up talking to Paul and Ron. Ron thought it was interesting and asked me to send him a follow up email. This is what I sent, I just dug it out and thought it might be interesting in this context. I've never really shared it with anyone either so I'd love some feedback even thought it's such a long time ago :)

--

OpenAnno is Twitter for Objects.

We want to build a close approximation of the real world "object graph".

OpenAnno stores public annotations (formerly tweets) on unique objects (formerly people). Objects are connected via typed edges (formerly mention, follow, reply), thus establishing the "object graph" (formerly "social graph"). Anybody can add annotations and edges.

Objects can be "followed" (an edge-type; there can be other types with other semantics). Objects can follow objects, too. A vendor can follow product ratings of products and adjust prices. This adjustment is pushed to following objects which might order those items once they are cheap enough. An alarm clock can follow your work schedule and your friends and toaster can follow the alarm clock. Thus, OpenAnno becomes the message bus in a global computational network. The interplay of smart objects reacting to one another is incredibly powerful (brains works like that, too).

Another application is search: the object graph can give you shopping advice that matches your wardrobe, can summarize and add value to complicated news events, can help you learn new languages and cultures, can suggest shops that sell every item in your fridge etc etc..

In certain domains (e.g. shopping) the object graph may be suitable to do analysis/research/simulation.

Possible "early annotators":

- Cars with license plates (a la bump.com)

- Anything with a barcode (a la stickybits.com)

- Real World that can be recognized in images (a la Google Goggles)

- Pictures/Media (comp. tazpic's presentation from yesterday)

- Objects other APIs give you

OpenAnno is a desirable partner for the mentioned companies allowing them to put their data in context. Data from non-partners is pulled in through "API tunneling" (as per yesterdays presentation).

Yesterday's demo is at http://www.justin.tv/jonashuckestein/b/264542955#r=Fs6GQLI~&... and a working prototype of the API at http://openanno.com. It's built to scale but a rewrite will be necessary (current code is public on github). Everything was built in less than 24 hours.

Challenge: We can build the "starting graph" from all the information we can gobble. This is difficult. Getting from there to a close approximation of the "object graph" is even more difficult and interesting.

Social proof: PG first called OpenAnno a "superset of Twitter annotations" and then a "communication channel that is bigger than Twitter".

--

sigh You gotta love the "it's built to scale but a rewrite will be necessary" line and the "social proof" section :) Makes me feel like I've come a long way in those two years.

Me and my friends talked to SV Angel for a little but I had already decided to take a different job for immigration reasons. Back then I didn't realize that these people were kind of a big deal and I should somehow capitalize on that. Good times :)

Edit: Formatting