The Browser Company of New York (BCNY), known for Arc, has just launched Dia through their alpha program – their latest take on the web browser.

Just as some background, BCNY created Arc in 2021, but late last year they put Arc on maintenance mode. At the core of this decision was that even though Arc had great traction, they needed something that would put it solidly ahead of Chrome – having a better UI wasn't enough to create a long term business. They decided to make an AI native browser – the result of which is Dia.

As an early Arc adopter and daily user, I was keen to take it for a spin.

This is a picture of what the UI looks like:

You have a sidebar on the right which can be triggered via cmd + shift + e, and the default new tab view is a search engine box that doubles as a chat interface.

The base model that you chat with is GPT 4.1, and the inline chat has the current page's content as context.

Features

Dia introduces several interesting ideas:

• The URL bar is neat: it pulls the query string for searches or displays the URL base and page title, rather than the full link.

• It intelligently routes queries. For example, if your query is phrased as a question (i.e. it begins with where, what, who, etc.), Dia routes it directly to its chat interface.
  • This is a smart move, as making chat the default for all searches (an issue I've found with trying to set Perplexity as a default search engine) would introduce unwelcome latency for basic searches.

It's got some small issues though:

• if you search 'werewolves', it still thinks it's a chat query.
• if you paste in a link, weirdly it tries to search it up instead of navigating to the site directly.

• The ability to toggle between the two modes is clever.

Issues

The spacing of the screen is a big issue:

Fullsize on a display (Macbook Pro) - the sidebar for Arc is roughly 1/5 of the screen.

While the AI chat sidebar takes up 3/10 of the screen real estate.

This really hurts usability for websites, as it creates an awkward dimension that many sites don't support well, leading to content reflow or truncation. Arc's sidebar doesn't play well with some sites to begin with, and having a sidebar almost double the size doesn't help. It's not Dia's fault, and you aren't always chatting with websites, but if this is a unique selling point for Dia, it better work well.

That said, the built-in adblocker is quite good – even better than uBlock Origin Lite.

On https://adblock.turtlecute.org/ Dia gets 77%, while uBlock Origin Lite gets 70%.

Conclusion

If BCNY's stated goal for Dia was to create a browser similar to Chrome for an easy switch, they've clearly succeeded – the resemblance is striking. The rationale makes sense. However, it's a tricky position. For the features where Dia aims to innovate and differ from Chrome, they really need to stand out. In other words, if you come for the king, you better not miss.

I think Google has stronger AI competencies and is better positioned to execute on these different agentic features. Consider a feature like agentic browsing: this requires multimodal capabilities that are not widely available and an infrastructure that can serve these AI features affordably enough to billions of users.

Personally speaking, Dia is missing the 'wow' factor I felt with Arc when I first used it—like its novel sidebar interactions and profile switching. I suspect those accustomed to quick profile switching in Arc will sorely miss it in Dia.

Some possible directions I can see this going:

The bear case:

• BCNY gives up on building Dia and instead figures out how to continue with Arc – this I feel is the most likely outcome.

The bull case:

• Google faces antitrust issues and/or is unable to execute as well on their AI-integrated browser, giving a significant advantage to Dia while they refine their offering.
• Dia finds a way to make users truly love the product (perhaps through exceptional agentic features or a standout design?).

In any case, I will always respect those willing to push all the casino chips in. BCNY is clearly doing that with Dia, and if they succeed, it will be a magical product.