Who are AI browsers for?

OpenAI launched an AI-powered web browser It is called ChatGPT Atlas This week, which makes me wonder: Is it finally time to ditch Safari?

Max Jeff, Sean O’Kane and I discussed the browser landscape – incl Some lesser-known options — in its latest episode equity podcast but none of us seem to be making a big switch anytime soon.

For one thing, Sean points out that many companies have tried and ultimately failed to break free of the major browsers due to their inability to make money on the browser alone. Of course, this is unlikely to be a problem for OpenAI Increasingly extensive financing rounds.

Meanwhile, Max has actually tried using Atlas and other browsers that promise AI agents will do the work for you, and he says there’s a “slight efficiency gain” at best. Other times, you end up seeing the agent “click to a website” — is that something normal users are really screaming for? Plus, there is Significant security risk

Read a preview of our conversation below, edited for length and clarity.

Anthony: I’m still on Safari, but as far as search engines, which are tied to browsers, I’m actually trying to experiment with non-Google. [options,] Because I’m tired of seeing all the genAI stuff at the top of my search results.

I think there’s also the question: If these AI browsers take off, what does that mean for the idea of ​​an open web in general? You can still go to web pages, but I don’t think it would be crazy to suggest that a website is just becoming less and less important as more and more of our browsing is controlled by these AI interfaces and chatbots.

Maximum: I think that’s a big idea that people talk about a lot: what does the agentic web look like? And I think that’s an interesting question. People have tried to come up with all these solutions to work towards this future [they] Feelings are coming

And I think there’s a certain aspect of it that reminds me of earlier technology waves where it’s like, “Okay, but what’s the actual experience? What’s the value proposition to a consumer to use one of these tools?”

And it’s not just super compelling today. I’ve tried ChatGPT Atlas and I’ve tried Comet and the most generous estimate between them is, it’s a slight efficiency gain. It makes you a little more efficient.

But most of the time when I’ve tried these things, you’re slowly watching it click around a website, doing something I’d probably never do in the real world. I want it to, like, look up a recipe and add all the ingredients to Instacart. I’ve never done it. I think all the tech bros always say that example in the videos, and I’m like, “I don’t know if people are doing that much.”

This is a huge gap facing the tech industry right now [saying] “We’re building all these tools for the agentic web,” but why would the average person use them? And I don’t know.

Sean: I haven’t used any of them [AI browsers] But that’s in large part because I’m still pretty old-fashioned when it comes to searching and browsing in general – a lot of the work I’m doing involves searching for documents, which naturally involves looking at various disjointed parts of web pages I’m familiar with, lots of Boolean searches on Google. Maybe I’ll try it one day if Google really stops boolean search, which seems to be coming at some point, but it’s not there yet.

What’s interesting to me about these AI browsers is that we’ve seen other companies try to compete in the browser space and they always lose because it’s impossible to make money on a browser as a product. And some have tried to charge for it up front, they might get a bit ahead for a while, but it’s ultimately unsustainable when it comes to competing against Safari or Chrome or Firefox.

What’s interesting to me is … you end up having these companies that have infinite money, so they can run it as long as they want, because they’re not actually trying to make money on these things yet. Eventually they probably will, but OpenAI doesn’t need to make money for this thing in the next year or two, they can take it out there and let it take shape.