The web remembers everything about itself and almost nothing about you.

Think about that for a second. Every website you visit knows exactly what it wants to show you. The one sided nature of the web has always been about labeling and grouping you into a certain bucket. But from your perspective, the person staring at a screen, the web is an endless stream of things that were never designed for you. They were designed at you.

We’ve gotten used to this. We open a browser, go to URLs, consume what’s there and move on. Maybe we bookmark something, or too manyu things...we’ve all been there. Maybe we forget to. Either way, the browser itself doesn’t care. It doesn’t get what you’re interested in, what you’ve been researching, or that you went down that rabbit hole about generative design systems or and never quite found the article that tied it all together. It also does not care if you prefer things more visual and less walls of text (as someone writting one wright now, I know some of you may bail when you finish reading this sentence). But maybe there’s a new way to think about content delivery, the nature of the web and web browsers in general.

The browser is one of the most used piece of software in the world (web views also count, ok). And it has no memory (just yet). We are starting to see movement towards the future of the web and software as a whole, but let me talk about the vision.

The gap most are missing

There’s a lot of conversation happening right now about AI agents, agentic interfaces, MCP, and the future of software. Most of it focuses on what AI can do: execute tasks, generate code, automate workflows. The discourse is dominated by capability. What can the model do? How many tools can it call? How autonomous can it be?

But there’s a gap in this conversation. A big one.

How can we use these new capabilities to provide entire new and more personal ways to think about and consume software. About software that changes completely based on who you are, what you care about, and how your interests evolve over time. Not something that is trying to categorize you to feed the same group, the same thing. But instead to actually help you get the most as it gets to know on your preferences, taste and personality.

There is a lot of focus on execution, not understanding. On tasks, not people.

I’ve spent years obsessing about web performance and talking about it. Talking about the convergence of product and performance metrics, what it means to serve the users the best and most delightful experiences. Caring that much about user’s experience taught me something deeper: the things that matter most to users are often invisible to the systems that serve them. The gap between what users experience and what systems optimize for is where the real problems live, and I’m not just talking about metrics here.

The same is true for personalization. Not the shallow kind "recommended for you" based on what you clicked yesterday. I mean something deeper. Something that actually knows and is there to serve you.

We see parts of this vision now taking shape as we starting to build this type of systems.

Why the browser

There’s a reason I keep coming back to the browser as the right home for this.

Every other platform is a walled garden, and even on the web there’s a rise of different closed systems. Each one optimizes for its own engagement metric and ignores everything else about you.

The browser sits at the intersection of all of it. It’s the one piece of software that touches every part of your digital life. And yet, for now, it’s just a viewport. This current model is changing and evolving but I feel like there’s a greater unrealized potential at hand.

I’ve once wrote that the browser should become "the last super-app." I think that’s directionally correct, but the framing is wrong. It’s not about cramming more features into a browser. It’s about making the browser actually aware of the person using it. More like an assistant that helps curate and surface things the way you like and for the things that trully speak to you.

That’s a different problem. And it’s the one worth solving.

What does it mean for a system to remember about you?

When I say memory, I don’t mean a database with your browsing history. I mean a system that understands the difference between a passing curiosity and a genuine interest. That can tell when you’ve been circling a topic for days versus when you stumbled on something once and moved on.

Consider how your own memory works. You don’t remember every article you’ve ever read. But you do remember the themes that keep pulling you back. The ideas that connect across different contexts. The feeling of recognition when you encounter something that fits into a pattern you’ve been building, maybe without even realizing it.

That’s the kind of memory layer I’m interested in building for apps now.

The problem is that this is genuinely hard and also as most things related to AI, we are still kinda figuring out. Same for security layers around those and how to best balance capability with privacy.

But one thing is for sure, systems that remember you and facts about you will be more enjoyable to use.

What comes next

I’m building something based on these ideas. A an app that remembers, that develops an understanding of your interests over time and uses that understanding to surface things you’d actually care about—without you having to ask.

This side-project has been my main obsession since early last year and it was being conceptualized and slowly developed over the course of last year as I experimented with PerfAgent.

There’s a lot more to say about how this works in practice: how interests are built, how the system understands your taste, how it presents things, how trust works when your browser starts acting on your behalf.

But those are topics for another post.

For now, I’ll leave you with this: the web was built to connect people to information, and other people. Somewhere along the way, it became very good at connecting information to people. The direction matters.|

And I think it’s time to flip it back.