What should we build?

a question that waited too long to follow "what can we build?"

Filed February 2022

Here’s the problem: The dominant paradigm of big data and AI in consumer tech so far has been extractive. Since 2013, even while we’ve made a diverse set of breakthroughs in CS research, a lot of work around applying big data and AI has gone into answering, “how can we extract more ad impressions out of each customer?”

It seems odd to me that this is how these innovations have been applied over the last 10 years, because a far more interesting and valuable question, in my view, would be, “how might we create new experiences that were never possible before?” and more specifically, “How might we create experiences that change your life for the better?”

The ads-based model is a trap. From a game theory perspective, it is teeming with powerful incentive misalignments. For example, under the ads model, the fundamental metric driving profit is more total ad conversions. This eventually means more ad impressions, which requires either cannibalizing the actual product and replacing it with ads—or it requires increasing total time spent in the product.

Then they make a shaky assumption to justify this business model: that people must only spend their time on things that are worth it to them. Therefore, if people are spending more time in the app, it must be good for them, too.

Here’s the obvious problem: except in rare cases, that assumption simply doesn’t hold up. We are human, and we’re all victims of our own irrationality. The time we spend on something is almost completely different than utility we get out of it.

It generally requires a lot of mental gymnastics to untangle the web of incentive misalignments.

And I’d rather not bother.

I hold a fundamentally different view: I believe we should create things that are so valuable, they’re worth paying for.

Let’s go one step further: I want to build products of supreme, unambiguous value. Where people get tons of real-world utility out of something with as little time required as possible.

This idea is fundamentally impossible for products in the attention economy—the singular goal of ad-based products is to consume your time, not create value for you.

I believe we should have products that let you use your data as a superpower—to help you do things you could never have done before.

And products should let you wield your data like a tool—to achieve goals that benefit you, not random third parties. Technology should support your life by stepping in at the right moment to provide value—and then it should melt into the background so you can do the things that matter, like investing in new relationships or celebrating old ones.

This is what we should build: products that give you superpowers, which unlock experiences you never could have had before, and focus on creating immense, and unambiguous, value for you directly.