An important thing to understand about Apple is it is a company that builds products for people to buy. This sounds like a very simple and obvious thing but in tech it’s actually quite rare – most tech companies build platforms, networks, services, or ecosystems. Apple, of course, has their own platforms, networks, services, and ecosystems but they all serve a product, preferably a beautiful device, or an app that runs exclusively on that device.

Apple News is a product that captures the interests of a person who has bought an Apple device. The very first thing it does is ask you what you like, in the form of newspapers, magazines, and blogs you already read, or topics you pick. It uses machine learning to refine its understanding of what you like and will, eventually, suggest new things for you to like1.

Just as important, the articles, stories, and posts in News look beautiful and designed, not merely dumped into a feed with an auto-layout engine. This will help ensure content that is not just relevant but high quality. If Apple can also deliver on the promise of letting publishers make money from News (a big ‘if’ considering not one single ad was demonstrated), they will have truly built an amazing news product that had mostly been written off.

Given Apple’s very public stance on privacy in opposition to Google and Facebook (cf. Dustin Curtis’s Privacy vs. User Experience), it raises the question about how good the machine learning (and ad targeting especially) can be in News. Apple believes they can do in-device what everyone else does in data centers, which seems like a reasonable enough approach until proven otherwise.

If News is successful, it means more than another good default app on Apple’s devices, it means tremendously useful data – a personalized Interest Graph2 – for other services, like Siri, possibly even available to developers. Then it becomes more than a product but a foundation.

Watching the demo reminded me of the very smart, though apparently not widely used, Prismatic. The people who built Prismatic are brilliant at constructing machine learning algorithms but don’t seem to have been able to turn that into a product. It’s perhaps worth noting that one of Prismatic’s co-founders worked at Apple until very recently. Prismatic also offers an Interest Graph API that I would not be surprised to learn Apple licenses. ↩︎

An “individual corpus of every user’s interests and passions” sounds familiar. ↩︎