None of the "things you can do" lend themselves to justify me having to host the micro-app - eg, they do not tell a story of needing to pull in additional application logic: was it a conscious decision not to build an IDE and hosting environment for human.io so I could just jump straight into building micro-apps?
We built the assembly language; the ide/hosting/etc would be a layer above that.
The events are pushed back to a webserver, and you decide what to do there. If you already have an application and data storage etc already, this is basically the easiest thing to do. We felt this flexibility would lead to the most interesting initial apps; additionally, the vast majority of early adopters would probably not love being sandboxed.
 we also have a hanging HTTP api; call /next_event and it hangs until it has events for you. this lets you talk to the system without having a public open port.