Burning the Browser Boats
The Browser Company
It's a good, well-articulated post from Josh Miller, the CEO of The Browser Company – notably, the makers of Arc, the newfangled Chromium-based web browser that has built a loyal user base over the past several years. Though not one large enough to achieve their ultimate goals. So they've started over with Dia, a new web browser built from the ground-up to be AI-first.
If this sounds familiar it's because Miller did a similar post – a video, actually – seven months ago. While they weren't quite ready to talk about the direction of Dia yet, it was pretty clear what it was going to be. And it was also pretty obvious what the ultimate outcome would be, even if Miller didn't want to admit it at the time: the end of Arc. As I wrote back then:
I hope that whatever path they go down for this new product – something, something AI, obviously – works out for them. Mainly because I want Arc to stick around and it won't unless the company behind it can find success – true business success. But I also hope whatever they do ties into Arc in some way because without that "synergy" – shudder – Arc at some point will either be neglected or shut down or spun out or sold off. That's just the way of things. So as an Arc user, we need this new product to feed into (or off of) Arc in some way or it's just hard to see how Arc lasts.
I have no doubt of their intentions with Arc. And while Pierce views Miller's video explanation quite cynically, watching it, I do not doubt that The Browser Company wants to keep Arc going *right now*. I'm just not sure they'll actually be able to in either the failure or success state of the company now that the focus has shifted.
And so here we are. If there's a problem with Miller's post today, it's that he's still equivocating. He won't just outright kill Arc even though that's what they clearly want to do. So instead, he's trying to crowdsource ideas for how best to keep it going, just not under the management of The Browser Company. Perhaps he just saw what Mozilla did to Pocket, and the immediate "WTF?" response from would-be acquirers, who had interest in buying the service but were apparently rebuffed by Mozilla. Maybe the situation is not too dissimilar here. And maybe this will pull some interest out of the woodwork.1
One reason why The Browser Company isn't just open-sourcing Arc, in-line with Chromium itself, is that the technology that underpins the way they build it, the Arc Development Kit (ADK), is the same tech they're using to build Dia. So I have no idea how to square that circle, other than perhaps some outside team working pro-bono to keep Arc updated with Chromium fixes.
But I'm not gonna lie: all of this makes me more hesitant to start using Dia. While I was one of the first testers of Arc, it took me a while to warm to it because several UI elements turned Chrome on its head. But once I did, I was hooked and it has made Chrome almost unusable to me. But now I may not have a choice but to go back – though the DoJ may have something to say about that too!
I've actually had access to Dia for a few weeks now – thanks, old trusty college email! – and it's nice in how simple it is. It's probably a little too simple for my taste right now, but I'm also unlikely to be the key demographic they're going after here because again, I was an Arc power user! I learned all the little tips and tricks that 99% of "regular" users won't, and Dia seems much more in line with a browser that the 99% will enjoy.
To that end, I enjoyed Miller's anecdote of some advice he got while building Arc:
First, simplicity over novelty. Early on, Scott Forstall told us Arc felt like a saxophone — powerful but hard to learn. Then he challenged us: make it a piano. Something anyone can sit down at and play. This is now the idea behind Dia: hide complexity behind familiar interfaces.
"Now that's a name I've not heard in a long time..." Yes, that's Scott Forstall, the former Apple executive who was instrumental in bringing to life a certain mobile OS that many of us all know and use today. And it's a fun analogy. That again, is probably the right advice if you want to reach true consumer scale.
I also appreciate Miller's thoughts on how best to bring forward an "AI browser":
New interfaces start from familiar ones. In this new world, two opposing forces are simultaneously true. How we all use computers is changing much faster (due to AI) than most people acknowledge. Yet at the same time, we’re much farther from completely abandoning our old ways than AI insiders give credit for. Cursor proved this thesis in the coding space: the breakthrough AI app of the past year was an (old) IDE — designed to be AI-native. OpenAI confirmed this theory when they bought Windsurf (another AI IDE), despite having Codex working quietly in the background. We believe AI browsers are next.
I'd just add that ChatGPT itself is piggybacking on one of the oldest UI paradigms: the chat interface. There's clearly something to the notion of making the new thing look like the old thing, as counterintuitive as it may seem. So as much as I loved, loved, loved Arc's pinch-to-summarize (on mobile), it's undoubtedly too clever and too hidden for that 99%.
We're clearly on the verge of a new wave of browsers. Dia just beat Comet, the Perplexity browser, to market (again, in limited Beta), but it's coming in hot. Undoubtedly, so is some sort of browser from OpenAI itself. Oh yes, and there's the new Chrome, with Gemini baked in, which seems wild given the scrutiny both Google and Chrome specifically is under. But it's also clearly something Google has to do to keep competing here. And then there are the AI products coming that seemingly won't need browsers – or at least not in the way we current think of them.
I just honestly don't know what I'm going to use now. Which is both annoying and exciting!
M.G. Siegler
M.G. Siegler
M.G. Siegler
1 In writing about the potential of another company acquiring Chrome -- something the DoJ wants, but is undoubtedly not going to happen -- it's wild how many different people pointed to The Browser Company, and specifically, Arc, as a good would-be acquirer. Of course, that would undoubtedly require tens of billions of dollars and even then, per many of Miller's points, I'm not sure they would even want to buy such a "legacy" product. Then again, could they really turn down the opportunity to buy up some 3B+ users?...