Are AI Browsers the Next Big AI Trend?

0:00Today on the AI Daily Brief, are AI browsers the next important AI trend? And before that, in the headlines, the rise of the agent marketplace. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI.

0:18Hello, friends. Happy Friday. Welcome back to another AI Daily Brief. Quick announcements before we dive in. First of all, thank you to today's sponsors, Blitzy, Superintelligent, and Agency.org. And to get an ad-free version of the show, go to patreon.com slash ai daily brief. Lastly, as I've been mentioning, at Super, as you know, we are auditing basically the entire world right now to help companies figure out their most applicable agent use cases. And it turns out that once we've audited them, they need help actually building those things. If you are a dev shop, particularly if you are handling high six or low seven figure deals and contracts of that magnitude, I would love to chat with you.

0:52Shoot me a note at nlw at bsuper.ai with agent builder in the title, and I will get back to you over the weekend and early next week. With that though, let's dive into some very significant AI trends. Welcome back to the AI Daily Brief Headlines Edition, all the daily AI news you need in around five minutes. We kick off today with a theme that I am very personally acquainted with, one of, I think, the big startup and tech themes of the moment, which is agent marketplaces. According to TechCrunch sources, AWS is launching an agent marketplace as early as next week in partnership with Anthropic.

1:24The reports are that the marketplace will be unveiled next Tuesday at the AWS Summit. Now, not much here is known, but it's presumed that the marketplace will allow startups to offer their agents directly to enterprises through AWS infrastructure. It is very clearly an agent gold rush right now, not just in terms of the agents themselves, but also in terms of agent infrastructure. Earlier in the week, Microsoft announced a partnership with Replit to integrate Vibe coding more closely into Azure infrastructure. Salesforce and ServiceNow have also been experimenting with their own marketplace offerings.

1:53And I think it's a pretty safe assumption at this point that any company that has a current marketplace of offerings for the enterprise is going to bring agents into that as well. We are still very early in figuring out how agent distribution and services should work. One of the challenges of agents is that they inherently require so much customization, especially when it comes to enterprise deployments. These are not off-the-shelf technologies. And although they will become more plug-and-play over time, even in their simplest version, they're more complex than most enterprise software of the past.

2:20The reaction to this so far is mostly a head nodding, yep, makes sense. AI engineer Moad writes, Smart move bundling Anthropic with AWS's massive enterprise reach. They're basically creating a distribution machine for cloud-powered agents while taking a cut of every sale. The real question, will enterprises actually buy AI agents from marketplaces, or is this just tech giants playing copycat? Nobody's proven this model works yet. To be honest, that question is one I've thought a lot about, and I'm not really sure. I think that the short answer is there are definitely going to be agent marketplaces.

2:50There will be lots and lots of very lightly customizable agents that especially when it comes to individual consumer users or even individual enterprise users, it will make sense to have that sort of marketplace access. This is the sort of app store for agents idea. When it comes to larger deployments, marketplaces for enterprises might be a useful browsing or gallery type of experience, but I don't think marketplaces that we have in mind, which is basically a model of a consumer marketplace are going to be the thing there. Now, part of our bet at Superintelligent is that something that is going to require a marketplace is all of the services and infrastructure that sit around agents.

3:24Companies are going to need people who have experience with agent deployment, setup, monitoring, etc. to help them get up and running, or at least they will decide that that's a shortcut rather than just figuring out for themselves. And that type of talent and skill set is extremely apt to be organized in a marketplace. In any case, frankly, I'm glad to see lots and lots of different shots on goal when it comes to this. What's clear is that there is immense desire to get agents into the hands of companies right now as fast as possible, and whatever helps that along, even if it ends up being just an intermediate model or something that helps us figure out the actual approach, I think is going to be valuable.

3:56Now, one more piece of news around Amazon and Anthropic. Amazon is reportedly considering another multi-billion dollar investment into the company. Amazon has already pumped$8 billion into funding into Anthropic, representing about half of their fundraising to date, But Anthropic has done nothing but get more competitive since then. Its centrality to the rise of coding agents has put them in a hugely advantaged position, and yet they are going to continue to need copious amounts of resources to compete. The story in the Financial Times is very kumbaya and all about how the companies are already working together to pitch customers, etc., etc.

4:28Next up today, a big market milestone. Earlier in the week, NVIDIA very temporarily jumped above$4 trillion in market cap, the first company to do so. Now, it has come down slightly since then, but I think it's very telling based on where we are that it was an AI company that hit that milestone for the first time. Some more substantive news out of NVIDIA, the company is preparing to launch a new AI chip for the Chinese market as CEO Jensen Huang sets off for another Beijing visit. The Financial Times reports that Blackwell-based chips designed to meet export controls are set for release as soon as September.

4:58The chips are a version of the RTX Pro 6000 GPU, which is a prosumer or workstation product. The US market version costs around$12 ,000 per unit, which is far below the roughly $70 ,000 price tag on the top of the range GB200 chips. In addition to being a much less performant GPU, the Chinese market versions will ship without high bandwidth memory or NVLink, which is NVIDIA's ultra-fast networking solution. These components have been a big part of the puzzle in creating large coherent training clusters that can integrate 100 ,000 chips. Now, the Financial Times reports that these chips are currently compliant with US export controls, but given how fast those goalposts can shift, it's not entirely clear what that actually means.

5:35Still, NVIDIA is sticking to their view that locking China out of US-made AI chips is the wrong strategy. In a statement, the company said, With the current export controls, we are effectively out of the China data center market, which is now served only by competitors such as Huawei. China has one of the largest populations of developers in the world, creating open-source foundation models and non-military applications used globally. While security is paramount, every one of those applications should run best on the US AI stack. The news comes as CEO Jensen Huang prepares to travel to Beijing next week.

6:03He will be in town to appear at the International Supply Chain Expo, which is one of the largest events of the year for China's tech and industrial community. FT Sources said he's also set to meet with top government officials. Jensen sits in a very interesting position. Not only is he right in the middle of the dominant tech trend of our time, he's also right in the middle of arguably the most important geopolitical struggle of the moment. NVIDIA has seen its Chinese market share drop from 95 % to 50 % this year as export controls tightened, and although Jensen has been defiant against White House policy to choke off AI access, he's also serving as something of an intermediary between Washington and Beijing.

6:36In fact, Huang met with President Trump on Thursday ahead of the China trip. No sources were willing to speak on background about the conversation, and neither the White House or NVIDIA put comments on the record. Some people speculate that this was about Jensen trying to get reassurance that the chips would be approved before he goes on that trip, but the more tinfoil hat theories have him play more of a diplomatic role. Speaking of chips, TSMC records another big jump in revenue as the AI chip booms. The chipmaker posted 39 % annualized revenue growth in the second quarter, outstripping expectations.

7:05Now TSMC sits upstream of NVIDIA and other chip designers in the supply chain, so many view its financials as a leading indicator for final demand. And current demand growth is almost entirely coming from AI. The numbers were soft in the mobile and consumer segments which still make up the bulk of TSMC's business, but in addition to skyrocketing demand from NVIDIA, TSMC is also servicing growing new orders from Intel as they begin to outsource production. James E. Thorne, the chief market strategist at Wellington Altus writes, AI spending keeps chugging along, so much for the Cassandras calling for the end of the AI boom.

7:38Indeed. That though is going to do it for today's AI Daily Brief Headlines edition. Next up, the main episode. This episode is brought to you by Blitzy. Now, I talk to a lot of technical and business leaders who are eager to implement cutting-edge AI, but instead of building competitive moats, their best engineers are stuck modernizing ancient codebases or updating frameworks just to keep the lights on. These projects, like migrating Java 17 to Java 21, often mean staffing a team for a year or more. And sure, copilots help, but we all know they hit context limits fast, especially on large legacy systems.

8:09Blitzy flips the script. Instead of engineers doing 80 % of the work, Blitzy's autonomous platform handles the heavy lifting, processing millions of lines of code and making 80 % of the required changes automatically. One major financial firm used Blitzy to modernize a 20 million line Java code base in just three and a half months, cutting 30 ,000 engineering hours and accelerating their entire roadmap. Email jack at blitzy.com with modernize in the subject line for prioritized onboarding. Visit blitzy.com today before your competitors do. Today's episode is brought to you by superintelligence, specifically agent readiness audits.

8:44Everyone is trying to figure out what agent use cases are going to be most impactful for their business, and the agent readiness audit is the fastest and best way to do that. We use voice agents to interview your leadership and team and process all of that information to provide an agent readiness score, a set of insights around that score, and a set of highly actionable recommendations on both organizational gaps and high-value agent use cases that you should pursue. Once you've figured out the right use cases, you can use our marketplace to find the right vendors and partners. And what it all adds up to is a faster, better agent strategy.

9:16Check it out at bsuper.ai or email agents at bsuper.ai to learn more.

9:46to solve complex problems. However, right now there is no standardized infrastructure for these agents to discover, communicate with, and work alongside one another. That's where Agency, spelled A-G-N-T-C-Y, comes in. Agency is an open source collective building the internet of agents, a global collaboration layer where AI agents can work together. It will connect systems across vendors and frameworks, solving the biggest problems of discovery, interoperability, and scalability for enterprises. With contributors like Cisco, Crew AI, LangChain, and MongoDB, Agency is breaking down silos and building the future of interoperable AI.

10:23Shape the future of enterprise innovation. Visit agency.org to explore use cases now. That's A-G-N-T-C-Y dot org. Welcome back to the AI Daily Brief. Today, we are discussing something that could either be a total nothing burger, A small little blip footnote in the development of the AI space, or could be extremely significant. Although it's very different, it has a little bit of the feel of the AI wearables pop that happened last year. Although I think that there might be reasons to be a little bit more optimistic just in terms of how well this trend matches the actual moment than we were back then.

11:02What I'm talking about, of course, is AI browsers. This week we had the launch of one, rumors of another, and a third in beta starting to get responses as well. Now, what this is all about and the context that makes this matter is, of course, the agentification of everything, including the very way we use the internet. We are speedrunning the move from theory to practice when it comes to agents. It was just months ago that we got Anthropics computer use mode, and shortly after that, OpenAI's first version of Operator, the Chinese agent Manus picked up some viral interest, and a common thread in all of these was that they were at core about deploying an agent to interact with the web and the internet in a way that you used to.

11:46In other words, there is a vector of agentic competition that is structurally reimagining how we interact with the web because instead of us interacting with the web, it is us deploying agents to interact with the web on our behalf. Now, one could argue that the rise of AI search, both in the form of AI overviews on Google, as well as just the shift away from traditional search behavior to trying to get answers in platforms like ChatGPT, is also part of this trend of deploying an agent or an AI or an LLM to do your internetting for you rather than doing your internetting yourself. But let's talk about what was actually announced this week, or what was rumored this week, and then we'll talk a little bit more about both the reactions and the implications.

12:28First of all, we got a report. As early as last fall, we started to get hints that OpenAI was considering taking Google on on their own territory with a browser. Back in November, the information reported that they had actually prototyped a browser and that they had started to explore different types of deals that could power search features for things like travel, food, and real estate. At the time, the information wrote, OpenAI could decide not to launch the browser, though earlier this year it hired two people who were instrumental in the development of Google's Chrome browser. Well, this week, we got a report from Reuters that OpenAI is in fact planning on releasing that browser in the coming weeks.

13:01TechCrunch writes, OpenAI's browser is said to use AI to rethink how users browse the web. Supposedly, the browser keeps some user interactions inside ChatGPT instead of linking out to websites. Reuters reports that OpenAI's browser may integrate Operator, the company's web browsing AI agent, as a key feature. So that was in the rumor column, but then we actually got a launch as well. Perplexity launched what they called Comet. In their announcement tweet, they said Comet is a browser that's designed to be a thought partner and assistant for every aspect of your digital life, work, and personal.

13:31Now lastly, I would be remiss if I didn't mention that the browser company, who had previously made Arc Browser, also has something similar called Dia that's in early access beta right now. These guys were technically first to market in June, but like I said, because it's currently behind beta, people haven't had as many reps with it yet. So let's look into Perplexity's comment to try to get a sense of what these types of browsers might actually include. Perplexity CEO Aravind Srinivas writes, Amit is a web browser built for today's internet. In the last 30 years, the internet has evolved from something we simply browse or search.

14:02The internet is where we live, work, and connect. It's also where we ask questions. Curious minds have questions everywhere, and they'd find answers on every page and every idea through every task. Yet we've been trapped in long lines of tabs and hyperlinks, disjointed experiences that interrupt our natural flow of thought. In other words, the internet has become humanity's extended mind, while our tools for using it remain primitive. Our interface for the web should be as fluid and responsive as human thought itself. We built Comet to let the internet do what it's been begging to do, to amplify our intelligence.

14:30Comet powers a shift from browsing to thinking. Tabs that piled up waiting for you to return now join one intelligent interface that understands how your mind works. context switching between dozens of applications sites and interfaces has stolen the focus and flow that bring joy to our work and fuel our curiosity comment transforms entire browsing sessions into single seamless interactions collapsing complex workflows into fluid conversations ask comment to book a meeting or send an email based on something you saw ask comment to buy something you forgot ask comment to brief you for your day comment transforms any web page into a portal of curiosity highlight any text to get instant explanations explore tangential ideas without losing your original context.

15:07Ask specialized questions or broad ones. Comet understands that genuine curiosity doesn't follow predetermined paths. With our own roadmap and with every new advancement in AI, we will continue to launch new features and functionality for Comet, improve experiences based on your feedback, and focus relentlessly, as we always have, on building accurate and trustworthy AI that fuels human curiosity. The future belongs to the people who never stop asking questions. A great mission statement, but what actually is this thing? Well, one either accurate or potentially reductive way to look at it, depending on your point of view, is to view this as a browser with an agentic assistant strapped onto it.

15:44Perplexity's version of the AI browser puts the agent in a sidebar, so you can call it up at any moment without switching windows. It can see all the tabs you're working in and take actions on any of them, so theoretically it can do anything you ask it to on the web. Perplexity demoed the agent, constructing a walking tour based on a map the user was looking at and generating a summary of a Slack window. Their browser also has voice mode enabled, so they showed a user pulling up a YouTube video based on verbal instructions. Now, none of the individual pieces then are all that new, but the change in form factor is what matters.

16:13Operator, for example, has done all of its web-based tasks in a cloud sandbox. That meant that workflow was very segmented between the user and the agent. It was mostly useful if you wanted to be able to complete a task end-to-end, rather than being able to seamlessly hand control back and forth between user and agent. Comet and these native AI browsers then seem far more suited to being able to complete a task when you're already halfway through or step in to provide some help. For example, it's easier to use Comet as an assistant to help draft an email and a web client. Now, another small but powerful feature is that Comet is designed with multi-agent architecture.

16:45That means it can fire off multiple agents to complete subtasks working in their own tabs in the background. Everything is also hosted natively in the browser, so Comet has full access to everything you're signed into during the session. All of this adds up to, theoretically, a feeling that isn't just an assistant bolted onto a browser, but an actually fully integrated experience, where the way that you browse the web involves having access to this intelligent Sherpa or guide that you can ask about anything that you're interacting with and actually have help you on whatever it is you need help with.

17:12Now, Dia, from what we know so far, is very much on the same page. They introduced a similar interface upgrade this week that they're calling inline browsing. CEO Josh Miller wrote, Now Dia opens web pages within your AI chat threads, blurring the lines between three categories of software, a web browser, a search engine, and AI chat. He continues, By building AI chat on top of a browser, not as a separate product, you get a more fluid thinking environment. AI and the web fuse together as one. I know that sounds pretentious, but it's the best way to describe the feeling of inline browsing. Web pages are just the beginning.

17:44Dia can render all kinds of interactive embeds in line. Today, that means you can consider purchases without leaving chat. Tomorrow, Toby's investment in MCP might mean shopping in line too with dynamic store embeds. A new internet is coming. Now, while we don't have access to OpenAI's forthcoming browser, I get that they're exploring pretty similar space. Now, right now, access to these tools is a little bit limited. Comet, for example, although generally available, is locked behind Perplexity's premium tier, which is$200 a month. So there are only a few people who played around with it. Still, among those who have, the response so far is pretty positive.

18:18Olivia Moore of Andreessen Horowitz posted, First test of Perplexity's new agentic browser, Comet. It pulled a list of all my email newsletters and unsubscribed from the specific ones I asked it to. In my opinion, this is much more useful than what we've seen from OpenAI's operator or even Google's Project Mariner, which is their own agentic browser and beta. The fact that it can actually do things for you within the application that hold all your data in context is really helpful. She continued some of my other favorite use cases, checked into an upcoming flight, It fills in confirm code, last name, etc.

18:45Looks for any unpained build in my inbox. Filtered my LinkedIn request by who I should accept. It's a tough grader. AI entrepreneur Nathan Snell writes, Holy crap, Perplexity's Comet browser is insane. Operator was a total dud. Manus is better, but meh. I asked it to duplicate a meta campaign for me. No problem. All automated. Running through some additional example use cases, he continued, Automated audit of Ridge's cart. Handling it like a champ. It's legit doing the things. Adding items to the cart. Seeing what changes. Brief break for dinner, but also Comet could totally order delivery to my house.

19:16If my wife and kids weren't already waiting for me, I'd do it. Honestly, this is what I had hoped operator was. Feels like the 4.0 image gen release. Kudos to perplexity. So again, if none of the exact use cases or features here are all that new to you, I don't think that's a wrong assessment. What matters is the idea of integration of agentic experiences as part of our core platform experiences. And clearly, these entrepreneurs are thinking about this as more than just a quote-unquote web browser. Back in March, Perplexity CEO Arvon Trinovast said that his goal with Comet was to quote develop an operating system with which you can do almost everything.

19:51In short, this is another step in the path towards redesigning the interface for a world where AI agents rather than humans are the ones interacting with computers. Matthew Berman of Future Forward referred to this as vibe browsing, which is certainly of the current zeitgeist. He said, I see a glimpse into the future in which web browsing is completely different. It's me tasking an agent to go and browse the web on my behalf. And in fact, I think that that's really helpful because part of the issue with thinking about this future is the term browsing the web itself. At this point, very little of our time on the internet is spent clicking around looking for content to read.

20:24Instead, the web has become the portal to a ton of administrative shopping and productivity tasks. AI browsers then aren't powerful because they can read and summarize websites for you. They're powerful, at least potentially, because they can act as generalized automation tools for web applications. Berman commented that by using an agent to interact with the web, quote, I'm going to be able to do so much more in parallel. I can kick off multiple agents. I can have scheduled agents. They can all run at the same time and just get so much more done. I don't want to go through the process of booking a flight.

20:52I don't want to put together groceries. I don't want to do a lot of things that I do on the internet every day. I want an agent to do it for me. Now, you've probably frequently heard me be skeptical of many of those use cases, but I'm certainly open to the idea that that calculus changes if they're just embedded obvious features of a more generalized experience like a browser or an operating system. Alex Gravely, one of the developers of Comet, wrote, Comet is the first big step in merging AGI into daily life, right from the search bar. Some people definitely get the significance and see how central to the AI battle this is going to be.