Your AI Doesn’t Know You (Yet)

If you’ve been anywhere near the internet this month, you’ve heard about OpenClaw, the open-source project that lets almost anyone create personal AI agents to manage their calendars, book flights, and really anything you can dream up.

People have been talking about personal AI, well, since we’ve been talking about AI. But this is the first time it’s felt like a real thing. And because the excitement is coming from this unfunded open-source project (and not some big marketing campaign), it feels like a genuine signal. Tech hasn’t felt this alive since the fail whale.

Yes, it’s messy. There are security concerns and general anxiety about everything going sideways. But it’s worth paying attention to the excitement. Here’s why.

I think we’re about to realize that what we’ve called “personalized” isn’t personal at all.

Spotify builds playlists based on listening clusters. Amazon recommends what people like you bought. Nike ID lets you pick colors. These are merchandising tricks that sort you into smaller and smaller buckets and call it personal. It’s segmentation dressed up as intimacy. Very little of it actually learns from you. You’re more complicated. You have moods and your preferences change, and you can’t even explain why. We’ve all accepted that we have to manage our own algorithms to get a decent experience. That’s why private listening sessions exist. Because we’ve lived with this version of “personalization” for thirty years, we’ve forgotten what the real thing would even feel like.

The real thing feels like a relationship. Think about a trainer who could tell you didn’t sleep well last night not because you told them, but because they’ve been paying attention. Those moments happen because someone listened, noticed patterns, and adapted. Those relationships have mostly been reserved for people who can afford them and are lucky enough to find someone who genuinely cares.

The digital era tried to democratize that. But it had to fake the very thing that made it work. Duolingo gives you the curriculum. Peloton gives you a class. But neither one notices when you’re struggling, or adjusts because you mentioned you slept badly. It’s well-orchestrated content. What is AI personal training really about? It hints at what’s missing in these traditional digital fitness approaches.

The products coming next will learn from you through the act of working together. You show up, interact, share what’s working and what isn’t, and the experience evolves because they remember.

A product that truly learns from you has to know the difference between “I don’t want anything sweet” and “I’m diabetic.” One is temporary. The other is permanent. Most products treat everything the same. They overwrite your experience based on the last thing you said. Netflix is a perfect example. Watch one true-crime documentary, and it expects you to watch another. Software has always been able to remember. What’s new is the judgment to know what to ignore.

The entire consumer internet was built on the assumption that people want more options and more control. More is better, right? Then we realized we had too many choices, so we had to mask them. We used AI to generate product recommendations and curated lists to ensure everyone kept shopping. It’s exhausting. Here’s the thing: most people don’t actually want more choice. They want to be understood well enough that they don’t have to choose.

This is the kind of interaction I’ve been thinking a lot about. We’re building Ray, an AI-powered personal trainer. (It’s ok if you roll your eyes, but trust me, this is different.) Ray speaks to you, listens to you, and even watches you work out. The best trainers take away all the decisions. They guide you through the workout and make the whole process enjoyable. It’s early. But every week, we see moments that make us stop and pay attention.

Someone tells Ray that jumping is too hard on their knees, and the next session doesn’t include jumping. Someone says they’re exhausted, and Ray dials back the intensity (but doesn’t rewrite the whole program, because tiredness is temporary). In the fitness world, where apps are just libraries of content, that’s remarkable.

And people tell Ray things they wouldn’t easily share with a human trainer. That they’re going through perimenopause. That they’re depressed. That they’re sick of feeling the way they feel and need to make a change. They share these things not because the technology tricked them into vulnerability, but because the cost of honesty is lower when you’re not worried about being judged, and because the product responds to what you share. Take Sarah, for instance, who found that an AI trainer helped her beat her spring motivation crash by adapting to her changing energy levels throughout the season.

When ChatGPT launched, everyone had the same fever dream: AI that truly knows you, fits into your life, and helps with everything. Then we got distracted. Now, three years later, the question is still: “Where’s the consumer AI?” This is it, just follow the energy.

Developers tinkering with their personal AI agents right now are using an early version of something that will eventually be as ordinary as pulling out your phone. That gap between their excitement and everyone else’s experience will close. When it does, we won’t call these things “AI.” They’ll be brands. Some of them will feel like companions. And we’ll call them by their names.

For thirty years, we built technology that gave people more choices. I think the next thirty will be about knowing people well enough that they don’t have to choose.