Sell a a nicer "native" client, either by wrapping the website in a custom browser like http://mailplaneapp.com/, or make a version that looks as close as possible to Exchange.
I disagree on the extensions market not being lucrative until a critical mass has been reached. The 'robot' feature is a killer - you can host your own bot that watches and performs operations on a given wave. There are going to be some very compelling bots built right out of the gate, that companies are going to be willing to pay for (think: 'HIPAA compliance bot').
Agree. If Robots can work on more than 1 wave at a time (eg manage/monitor input on one wave and send summary updates to another) then it also opens up possibilities for lazy distributed processing applications. A (superficial) look at their proposals so far gives me the impression that the robots will make great quick-n-dirty FSMs and DSPs. There are many commercial opportunities here, but I also think think there's massive potential for pro bono and free-not-as-in-beer projects.
Robots can work on as many waves as you like. Minimally, a robot can just be a participant in any wave you want it to be a part of. You could technically have 5 Eliza robots in a wave, all talking to each other -- although that would probably be a very boring conversation.
There are others beyond this that go deeper -- with the open sourcing of the server and client you can imagine there will be applications that can be built that just take advantage of the raw capabilities of the wave server itself. I have one in mind, but I am keeping quiet until I have something to show for it. :) (On that, anyone want to hook me up with a sandbox account?)
Sell a a nicer "native" client, either by wrapping the website in a custom browser like http://mailplaneapp.com/, or make a version that looks as close as possible to Exchange.