@joshpuckett Awesome! We train a model per app. It doesn't make sense to invite your family to YesGraph like it does for Dropbox. You rank by last name, but you don't dedupe, so my recommendations are all stale emails of people that already have accounts. I just took this screenshot

@joshpuckett Solid advice! Our model considers that kind of data, and determines how useful it is for each app. For example, it might suggest lots of family members for a photo sharing app, but none for a dating app. The best part is that it gets smarter over time, so the data starts to speak for itself. It's definitely an interesting problem to be working on! Thanks for chiming in :)

@joshpuckett That said, my contacts might be atypical. Which is why we train across all an app's users to see what patterns work.

@joshpuckett Also, you may know @ikirigin from his Dropbox days?!

@carolynjoneslee @joshpuckett I don't think we overlapped :-D

@covrter Yeah, that was the design from the start, which meant a lot of changes on the tech side. For example, we were able to change how we do authorization in our API, without sacrificing security. I think we ended up with a much better product because of the mandate to make it “one line of code"!

@bradenhamm This flow is for when you want a user to share a link and send a message. What happens when the link is clicked and the invitee converts is up to you.

@_andrewlee We train a model using machine learning based on past invites. The contacts we recommend here are for YesGraph's model, which doesn't have as much data as our B2C customers. We just launched :)

@swetzequity Thanks! Our goal was to dramatically simplify adding YesGraph to your app