This is a rough draft; I haven’t given it much in the way of polish, and it kind of just trails off. But a friend of mine asked me what I think web browsers look like in 2025 and I promised I’d let that percolate for a bit and then tell him, so here we go. For whatever disclaimers like this are worth, I don’t have my hands on any of the product direction levers here, and as far as the orgchart’s concerned I am a leaf in the wind. This is just my own speculation.
I’m a big believer in Conway’s Law, but not in the sense that I’ve heard most people talk about it. I say “most people”, like I’m the lone heretic of some secret cabal that convenes once a month to discuss a jokey fifty year old observation about software architecture, I get that, but for now just play along. Maybe I am? If I am, and I’m not saying one way or another, between you and me we’d have an amazing secret handshake.
So: Conway’s Law isn’t anything fancier than the observation that software is a collaborative effort, so the shape of large piece of software will end up looking a lot like the orgchart or communication channels of the people building it; this emerges naturally from the need to communicate and coordinate efforts between teams.
My particular heresy here is that I don’t think Conway’s Law needs to be the curse it’s made out to be. Communication will never not be expensive, but it’s also a subset of interaction. So if you look at how the nature of people’s interactions with and expectations from a communication channel are changing, you can use it as a kind of oracle to predict what the next evolutionary step of a product should look like.
At the highest level, some 23 years after Netscape Navigator 1.0 came out, the way we interact with a browser is pretty much the same as it ever was; we open it, poke around it and close it. Sure, we poke around a lot more things, and they’re way cooler and have a lot more people on far end of them but… for the most part, that’s it.
That was all that you could do in the 90’s, because that’s pretty much all that interacting with the web of the 90’s could let you do. The nature of the Web has changed profoundly since then, and like I’ve said before, the web is everywhere and in everything now. But despite that, and the fact that browsers are very different beasts now than they were when the Web was taking its first tentative steps, that high-level interaction model has stayed pretty much the same.
But if the web is everywhere and in everything, then an interaction that involves opening an app, looking through it and closing it again seems incredibly antiquated, like you’re looking out a porthole in the side of a steamship. Even the name is telling: you don’t “browse” the web anymore. You engage with it, you interact with it, and with people, groups and businesses through it.
Another way to say that is the next generation of web browser won’t look like a browser at all: it will be a service.
More specifically I think the next generation of what we currently call a web browser will be a hybrid web-access service; like the current Web, it lives partly on a machine somewhere and partly on whatever device or devices are next to you, and act as the intermediary – the user agent – that keeps you connected you to this modern, always-on Web.
The app model is almost, kind-of-partway there, but in so many ways it makes life more complicated and less interesting than it needs to be. For the most part, apps only ever want to connect you to one place or set of people. Maybe that’s fine and that’s where your people are. But maybe you have to juggle a bunch of different communities in your life across a bunch of apps that go out of their way to keep those communities from discovering each other, and they all seem to want different slices of your life, your time and data depending on what the ad revenue people think is trendy this week. And because companies want to cover their bases you end up with these strange brands-pretending-to-be-people everywhere. It’s a mess, and having to juggle a bunch of different apps and communities doesn’t make a ton of sense when we’ve already got a reliable way of shipping safe, powerful software on demand.
I think the right – and probably next – thing is to push that complexity away from their device, to this user-agent-as-a-service living out there on a serverin the cloud somewhere, just sitting there patiently paying attention. Notifications – a superset of messaging, and the other part of this picture – can come from anywhere and be anything, because internet, but your Agent can decide whether forward them on directly, filter or bounce them, as you like. And if you decide to go out there and get something – a video, a file, a page, whatever, then your Agent can do all sorts of interesting work for you in-flight. Maybe you want ad filtering, maybe you paid for an antivirus service to give that file a once-over, maybe your employer has security protocols in place to add X or strip out Y. There’s lots of room there for competing notification services, agent providers and in-agent services, a marketplace of ideas-that-are-also-machines.
There’s a couple of things that browsers, for all their warts and dated ideas, do better than any app or monolithic service; most of those have to do with user intent, the desire for safety and privacy, but also the desires for novelty, variety and unique humanity. I’ve talked about this before, the idea of engineering freedom in depth. I still think it’s possible to build human-facing systems that can – without compromise – mitigate the possibility of harm, and mount a positive defense of the scope of the possible. And I think maybe this is one way to do that.
(Updated: Typos, phrasing, added some links.)