The magic of AI is in personalizing experiences to your unique context, but right now its so laborious to upload our context and life stories to each AI tool we use. In the future, we’ll have a sort of AI agent that has our full life context–our preferences, who we know, our life stories and experiences, and how we like to think–and this will communicate with all of our tools and products for us. That way, when we try our fourth AI calendar app, we don’t need to tell it that we don’t like meetings on Tuesday mornings because thats when street cleaning is. Or, we’ll never need to tell a food delivery agent that we’re allergic to nuts and prefer vegetables with every meal. We’ll tell our context agent once, and it will communicate with all your AI minions and real world products for you. I call this the Contextual Twin.
I spend a lot of time at the frontier, looking for new patterns where we can still invent and push things forward. Lately, it’s been the first generation of consumer AI products. I’m trying to play with it, to mold it, and to see how it stretches. What are the key characteristics of this new material? What happens when multiple people play with it? What characteristics make this material unique to build with? One of the key ones is personalization, if I give it enough context, I feel like the AI can see me, know me, and make things specifically for me. My favorite experiences with AI are when it has as much context of me as possible, and is able to use it intelligently.
Think about an AI scheduling assistant. It needs to know when you wake up, whether you prefer in-person or remote meetings, if you’re adventurous about new cafes or value routine. Now multiply that by every product you use. Word processors, travel agents, search engines – there’s no world where we spend half an hour explaining our decision-making process to each one.
Much like clay hardens when fired, AI becomes valuable only when you enrich it. Otherwise, it is just a helpful search engine. However, enriching our AI apps with data each time is a terrible user experience. What we need is some sort of shared identity + context that is portable and that we can use between apps. Much like a Google login. Imagine having to recreate every calendar event every time you wanted to try a new calendar app. That’s the current state of AI: very general, doesn’t get to use the material properties that make it magical, and you’re speaking individually with every single one.
In the limit, we are all going to have this sort of Contextual Twin–An agent that manages your identity and Whole Life Context. You will feed it just once and then passively over time, and it will know all the moments and ideas that make you, all the important people in your life, what you think and care about, you preferences and a record of everything you hate. And it will use that to tailor the world around you. The ideal case is an agent that knows every thought and experience you’ve ever had, transacting the relevant parts in a privacy centric way with every piece of software you use on your behalf. It’s like having an assistant go to the restaurant before you and tell the chef about every meal you’ve ever had and how that should inform what he cooks. In the age of AI, we allow the world to tailor itself to us.
This is the next frontier within computing + HCI: not just smarter software, but software that truly knows you. Software that makes you feel human and unique, software that caters the world around you to you. We’ve spent decades adapting ourselves to computers. Now, finally, they’re going to adapt to us.