hello delight

Taylor Davidson (@tdavidson) at Unstructured Ventures has written quite a bit about the concept of a personal API as a way to scale yourself.  I love this concept and thought I'd pick it up today.

If we simplify the concept of a personal API greatly, we can say there are two basic functions the API has to be able to do:
1) Request information
2) Respond to information

The former is possibly where the most research has been placed to date.  I equate this with something like the reverse of the Netflix or Amazon recommendation engines, albeit on a small scale.  The idea is that I have a personal API that disaggregates the signal from the noise in digital streams of data.  Presumably, this API would know the properties of what makes data meaningful to me and then seek them out.  The end result of this half of the API is still limited by scale as it takes me to process the data and refine the API to seek out better, more concise data.

The latter is most interesting.  The concept is not unlikely HAL.  I could program an API that would respond to inquiries as though it were really me.  I suppose there are early incarnations but I'm not sure where.  One area that may be suitable for discovery with Respond APIs could be in the auction market.  The variables to an auction are fairly few yet very personal.  In its most simple form, the API need only concern itself with reservation prices and an index of utility.

Where personal APIs get interesting is when the former and latter type begin to interact in order to "scale doing" as Taylor puts it.  Using the simple auction API to respond to auction inquiries, you can see how the data consumed by the request API could inform my reservation price as well as the utility of the item.  This is roughly the concept used by automated program trading systems, I expect, although these aren't used on a personal level that I know of.  The personal API has to account for emotion and irrationality, two very difficult concepts to model.

One distinction that I think is important to note is that a personal API is not a passive listening agent, in my opinion.  It's an active agent that I configure.  It's almost a manifestation of the scientific method or a decision-modeling theory.  But, it's not Bing and it's not Kosmix.  It's a way to scale "my" decision making and resulting action in an active way - not a way to simply filter out irrelevant messages.