✉️ To SigmaOS’ CEO: This Is What I Don’t Like About Arc’s Direction

I really, really like the Arc browser. But as I alluded to in this post, I have some reservations regarding it, and don’t feel like it’s going in a direction that I like. In the post, I said that I might try SigmaOS again — and I am. ¹

I mentioned this in their community Slack, and their CEO, Mahyad, asked me what about Arc’s direction I don’t like. I must say, the dev team seems very active, nice, and open to input! So this post is my reply to his question.

(And here’s a link straight to the TL;DR at the bottom.)

Hi, Mahyad — and thanks for asking! I wrote a blog post called «I Just Want A Nice Browser!», which might give you a hint, heh.

And let me also say that I’m a bit worried about your direction as well — but I’ll come back to that. 😉

Two fundamentals I don’t love, but that I don’t need to go too much into

1. I don’t love that Arc is built on Chromium — as I think Google has more than enough power over the web as it is.
2. I’m not against supporting any VC funded company — but in combination with an unclear business model, I become more skeptical and worried if our incentives align. ²

My main issue, though, is regarding AI

3. Focus

From my blog post linked above, in reaction to Arc’s launch of Arc II:

This is the moment I felt the Browser Company and I started to drift apart.

The entire premise of Arc II, is that «browsing the web sucks», so they would like to do it for me. But I don’t agree: I like to browse the web — I just want a nice browser to do it in!

Now, Arc is still a nice browser — and they haven’t stopped shipping new features to make it even nicer. But that direction just isn’t for me. And when dev resources aren’t infinite, focusing on big areas that aren’t for me, is a turn-off. ³

I like small features that enhance the product, like Instant Open. However, and I know I might be in the minority here, I hate Browse for Me. I don’t want the browser to browse for me, pick my sources, and keep me away from what, I think, might be my favourite tech service: Kagi Search. I want a more human web — not less. ⁴

Now, I do think LLMs, and other modes of generative AI, can be immensely useful. But the market is flooded with ways to interact with them. My personal favourite, is Raycast: Hyper + P gives easy access to a chat, regardless of which app I’m in, and I’ve made some small GPTs to help me with repetitive tasks (like checking how a word is spelled).

Sometimes having the AI know more about the context is useful. For instance, the one in Zed knows my code base. But it has to be quite a bit more useful than just using Raycast, for it not to be wasted dev time that could’ve been used to improve the core app instead! ⁵ It looks like you have some neat ideas with Aieris that leverages that it knows the web page — and I’m getting access to Airis today, to check for myself. ⁶ But my fear is that most of the features step so close to Raycast that it, for me, do end up being wasted dev time.

Because while the market for «general access to LLMs» is extremely flooded, the market for «nice web browsers» is far from it. Especially if you, like me, want to stay away from Chromium.

In my mind, there are two different ways companies seem to leverage AI these days:

1. Trying to be The One Big Portal To Generative AI.
2. To improve the details of their core product.

As I really like Raycast as number 1, personally, I want most other products I use to focus on number 2. Arc Max features that fit the latter, are things like Instant Open, Tidy Tab Titles and Tidy Downloads. ⁷

Some other (number 2) ideas for SigmaOS, from the top of my head:

• Automatically sorting tabs as subpages might be a bit too much. But maybe you could have the AI «sense» that a tab might belong under another tab or in another workspace, and then display a button? And if you click it, it sends the tab there (hover would show where).
  • There could also be a button next to the tornado, to do it to every tab. (Need undo for this, though!)
• Another use for the same feature, could be in the Mini-window: It could suggest a Workspace to open it in, based on the content. (And if it’s too unsure, it would revert to the default.)
• Let’s say you have some text on a website, and you want to search for that somewhere else. Maybe you could select it, hit a hotkey, and it would propose (three?) places to search for it. You should be able to say which services you prefer to search on, and perhaps pin some. (I might pin Kagi, so that that’s always suggested.)
  • If I select a video game title, it knows that it’s a video game and that I prefer to buy on Humble Bundle or GOG — so it offers to search on Kagi (pinned) and those two.
  • Now, some might prefer just asking «is this available on GOG?», but I like to actually go to the websites instead, and that the browser makes that easier. ⁸
• A variant could be that when you select text and hit a hotkey, you write something in a text box, and it acts on those two data points. Examples:
  • I select «Balatro» and type «Humble» = It searches for Balatro on Humble Bundle.
  • I select «Apple’s DMA compliance» and type «John Gruber», and it searcher Kagi for «Apple DMA Daring Fireball» (or something). The challenge here, of course, is knowing when a name is «a site to search on» and not.
  • The other way around, would do the same: Selecting «John Gruber» and typing «Apple DMA».
  • I select «SigmaOS» and types «YouTube», and it searches for videos there.

• Lastly, yet another variant, could be to use it to enhance the Lazy Search: On Kagi, I can use bangs to search on specific sites, like «yt SigmaOS» searches for SigmaOS on YouTube. ⁹ However, I have to write exactly «yt + string». It would be cool, if all of these, typed in Lazy Search, searched on YouTube «youtube sigmaos», «sigmaos youtube», «sigmaos on youtube», etc.

4. Ethics

Even though I, personally, find generative AI useful, I’m also very skeptical regarding the ethics behind them and if they’re a net good for the web and society. Too many people think that «If you think it’s useful, you must think it’s ethical» or «If you think it’s unethical, you must think it’s useless».

John Voorhees and Robb Knight make some great points in the latest episode of the Ruminate podcast (Full of Spicy Takes), and Matt Birchler also says stuff I agree with here.

These models are worthless without other people’s content and hard work, but become very valuable with it. Then it simply feels wrong that they can take this content without it benefiting the creators. And «no», I don’t care whether it’s technically legal or not.

Personally, I’d love to use image generators for my blog posts — but I’ve landed on that being too unethical for me. But as mentioned, I do use LLMs — but I’m also unsure about that. I do try to be mindful about how I use it, though. ¹⁰ And I want the implementations of tools to be as positive as possible — both for the web and the greater world and people it in. I don’t want them to make it too easy for practices like gunking up search results and platforms, or ruin the livelihood of creators. And I think that Arc (and things like Perplexity) are doing too much of the latter.

Things like Browse for Me, are trying to replace me going to the places where it got the information. And, I mean, you have to be pretty naive to think that this 5-Second Preview feature benefits Our Escape Clause:

On Hard Fork, the CEO of Perplexity said that they were trying to cite more sources, and that if your name often came up as their sources, that would increase your credibility and recognition. How that, in turn, would pay for your mortgage (when Perplexity just gives users all your content), he didn’t mention.

I wholeheartedly belive that stuff like this, if left unchecked, will drive content creators out of business and drive a large portion of the content that’s left behind paywalls. And both will hurt the web.

So, while I disagree with Arc, in that «browsing the web sucks» — their products are starting to do their part to make it more true.

TL;DR:

So, I already didn’t like that Arc was built on Chromium and was VC funded without a business model.

But, from a user perspective, I don’t like that they’re moving the product to focus on stuff I don’t want — and I think that focus has questionable ethics and is hurting the web.

Please don’t do this, Mahyad. 👆🏻 If you’re going to implement AI into SigmaOS, do it in a way that doesn’t remove focus from it being a nice web browser for actually browsing the web. ¹¹ And don’t do it in a way that hurts the real people making content and making the web a great place.