205 | AI browsers declare all-out war on Google, Apple eyes a $14 B search coup, Microsoft saves $500 M by firing humans — plus other must-know AI stories for the week ending on July 11, 2025

0:00Hello and welcome to a weekend news episode of the Leveraging AI podcast, the podcast that shares practical ethical ways to leverage AI to improve efficiency, grow your business, grow your business, your business, your business, your business. And before we get started, I would like to remind you that we are just a month away from the next public session of the AI Business Transformation course, which is a course that I have been teaching personally to thousands of people by now since April of 2023. The course gets updated every single session. We're teaching it at least once a month, but most of these sessions are private.

0:31They are for specific organizations or companies who book me to come and teach their people. and we only open courses to the public once a quarter. So the previous course was in May and this coming course is going to start on August 11th. So if you are looking for ways to improve your AI skills in order to accelerate your career or keep your business running, I highly recommend that you join this course. You can check out the link in the show notes to get more information about it. And you can use promo code leveragingai100 to get$100 off the price of the course. On that page with the link, you'll be able to see the entire syllabus and testimonials of what people are saying about the course and so on and come and join us a month from now.

1:11And now to the news. There is a lot to cover. There's some big interesting news. We're going to talk about AI browsers and their potential impact on the world as we know it. And we are going to cover some rapid fire items as always. So let's go.

1:28the first topic of the day is AI best browsers perplexity just announced comet which is their AI powered web browser they released it on July 9th you probably don't have access to it it currently only is available to their top premium users under the perplexity max plan that is 200 hours a month and beyond that it's by invitation only and that's probably going to expand over time. We're going to talk about why in a minute. But the idea is to integrate traditional search in a web browser environment with AI assistant that can help you do these things more effectively. There is a sidebar assistant that understands all the content in the web page that can browse across web pages that you can ask questions.

2:09It can summarize things for you. It can compare products for you and it can fetch additional information and complete things for you like complete forms, make bookings in restaurants and so on, including switching tabs and everything else we do as humans in a browser. Now it's built on Chromium, which is the open source version that Chrome is built on top. And the benefit of that is all your Chrome extensions are going to work. So if you are like me and you have 50 different Chrome extensions doing different things, they're also going to work in Comet. And so you're not going to lose any of your regular working environment if you are usually using Chrome.

2:42Now, the other thing that it's doing is that it's offering local data storage, which can reduce the amount of data that needs to go back and forth, and it can save you tokens in the process. It also has a native ad blocker, and it looks and sounds like a serious contender that can actually provide a lot of value. If you want the exact quote from Perplexity spokesperson, they said, and I'm quoting, we built comment to let the internet do what it has been begging to do, to amplify our intelligence. Now, the next company that has announced a AI browser or hasn't announced yet, but the rumors are getting very, very solid, is OpenAI.

3:15So OpenAI and I shared that with you in the past few weeks, are planning to release an AI-powered web browser. And it seems that it's just around the corner. It is aiming to go after Google Chrome, but different than perplexity, this is a much bigger play. And the reason it's a much bigger play is because, A, OpenAI is going straight after Google's core business, which is their ad business, with 500 to 800 million weekly users. If ChatGPT releases a browser, it is going to take a significant chunk of traffic from Google. And this is in a completely different scale than what Perplexity is doing right now, unless Apple buys them, which we're going to get to later on in this episode.

3:56Now, similar to Comet, it is going to be built on Chromium, which means the same thing. You'll probably be able to use all your Chrome extensions in that browser. And it's going to be built with an AI agent and web operator built into the platform itself, meaning it will be able to take action for you, browse for you, search for you, compare for you, and do everything that AI knows how to do, but in a browser environment. So let's go back to 30 ,000 feet for a second. First of all, from Google's perspective, their ad revenue adds up to about 75 % of all of Alphabet's income, 75%. It depends on the knowledge of what people are searching for so they can serve the best ads to them.

4:36From OpenAI's perspective, if they can get that information. First of all, they might find different ways to monetize what they're doing right now. But in addition, they will be able to get data on what people are searching for and the answers and the context of what people are doing, which will allow them to train their future models to be even better for these kind of tasks. Now, are we willing to let AI train on the data of everything that we search? Time will tell. I'm willing to bet that if it will provide us better results and more contextualized things and so on, while maintaining our privacy and anonymizing the information, over time, this will become very common and every browser will basically do that.

5:13And so we won't have much of a choice. Now, to tell you how long this has been in the making, in mid-2024, OpenAI recruited two former Google VPs from the original Chrome team that actually built Chrome as we know it today in the early days with the plan of doing something like this. So they've been at this for a while. They've been rumors about this for a very long time. And it seems to be that it's just around the corner and we're going to get access to a chat GPT slash OpenAI based browser that will integrate agents straight into this. Now, to expand this even beyond something we discussed several times on the show, there is a Department of Justice ruling against Google's monopoly.

5:52And one of the things they're trying to force them to do is to sell Chrome. And OpenAI raised their hand and said, if that ever happens, we will obviously be interested. And the reason is very obvious. It's the largest browser in the world. They get about 60 % of global browser traffic. And if you get your hands on that, you get a lot of information and a lot of distribution to many, many clients. But for now, since it's not on the shelf yet, because there's zero doubt in my mind that this will get to the Supreme Court before anything actually happens, then the next best thing from OpenAI's perspective that already has a very significant distribution is to develop their own browser.

6:24So beyond the fact that it's an immediate real threat on Google's dominance, what else does it mean? The first thing that it means is that it's going to change everything we know about search. And obviously it's not just open AI, it's a clear direction that all the companies are going to go to. So it's going to be open AI perplexity, and later on, probably Safari and so on. So all the big browsers will become AI browsers. And there are already some small companies with AI based browsers. But why does it change everything that we know about search? Because it means that as users, we're going to see less results and more answers, meaning instead of seeing a list of websites, we'll get answers to the questions that we're asking, which means less and less traffic to the actual websites, which over time will come to probably very close to zero.

7:07That over time can be three, five, seven years, probably somewhere in that timeframe. If your website, if your business depends on web traffic the way you know it today, you better start thinking how to change everything that you know, including the architecture of your website, including what you sell, including how you sell it to this new model. The problem is it's not very clear which direction this is going. But what is very, very clear is that the structure of your data and making it easily accessible and digestible to agents and AI tools is going to become significantly more important than the user interface and the UI that we worked so hard in order to optimize in the last 20 years.

7:43It also means that different scare tactics and scarcity tactics in FOMO are going to become less relevant, at least in the beginning, because whatever you put on your website, people won't see. They will only see the data that comes from the agent or the AI tool that they're using in order to get to the data. So all these things that are critical aspect of every e-commerce company in the world, and many companies are not e-commerce, are going to change dramatically. And we will have to adapt. And those who will figure it out quickly have a higher chances of capturing market share, at least in the short term.

8:16Now that raises a few bigger questions. One is what exactly are you supposed to do as a company? The one thing that is very, very clear is develop new channels of capturing an audience and capturing clients beyond straight web traffic. You could have invested the last 20 years in SEO and have grown a very substantial business based on that. All of that may go away. It's not going to happen overnight, but you better develop other channels. From my perspective, I'm really, really happy that a very significant part of my traffic and my business comes from this podcast and from my YouTube channel and from LinkedIn.

8:47So it's a lot more diversified than just the multiply.ai website. But if you are not in that situation, you better start developing these kind of additional channels because the traffic channel is definitely not guaranteed. Where exactly and how it will evolve is a different question. Will you be able to capture that because of that is a different question. But it's very clear it's not going to be what it is right now. The other aspect that you need to think about, or we all need to think about, is what is going to be the incentive of creating new content. Web 2.0, the introduction of everybody can share content, created a very strong incentive to create new content.

9:20Google SEO provided another incentive to create new content. If these agents work in a different way, maybe there is not going to be an incentive to create new content. And then if new content is not created, there's no new data to train on and there's a lot of other implications. Why would you share your information if it doesn't drive traffic to your website? There is a lot of other aspects in there that are not solved at this point, but it means we may have a very, very different web. But a different web also means maybe a different architecture and infrastructure for the actual internet itself.

9:48may be the way the internet works today, the HTTP and HTTPS protocols, and the way the physical internet is set is not optimal for the AI future. So maybe instead of internet, we need AI net, or Skynet maybe is the right solution to go with. I'm obviously joking or not completely joking. I'm not 100 % sure. I hope I'm joking. But the reality is that a lot of the things that we take for granted that became an integral part of our day to day might dramatically change in the next few years. And from open browsers to the big release of the week, which is xAI just launched Grok4, which is its latest flagship model.

10:23And it's scoring very, very high on many benchmarks above every other AI platform out there. The most notable one, because a lot of the ones are niche and nobody knows exactly what they mean, but the humanities last exam, which is a mix of multiple really high level, complex PhD level kind of questions in multiple topics that is currently the benchmark for really sophisticated, complex answers, Grok4 scored a 25.4 % on that test, which sounds very, very low. But when you consider the fact that the leader so far was at 21.6%, and that was Gemini 2.5 Pro, just ahead of OpenAI's 03 with 21 % flat.

11:02So it's a big jump from those. Now, in addition, Grok also introduced Grok4 Heavy. Grok4 Heavy is actually a very interesting architecture, which takes multiple agents that work as a team to research and debate information before they provide you answers. And if you want to use Grok4Heavy, you will have to pay$300 a month. But one of the things that Grok4Heavy has is the ability to use multiple tools. And Grok4Heavy with tools scored 44.4 % on humanity's last exam compared to the best score with tools that Gemini 2.5 Pro was able to achieve, which is 26.9. So it's a very big jump forward in its ability to solve really complex problems.

11:43So this is from a benchmark, what Musk himself said in the release. He said, with respect to academic questions, Grok 4 is better than PhD level in every subject, no exceptions. Now that's a very bold statement. On the other hand, it's Elon Musk who makes bold statements every single day. So how much is it accurate or not? I'm not 100 % sure. But what is very obvious that it's a very, very capable model when it comes to thinking, reasoning, and data that it has access to, and its ability to use external tools to do even better. What he also said is that it lacks common sense and that it hasn't yet driven any scientific breakthroughs.

12:18But he continued that by saying that it might discover new physics next year. Let that sink in. So the idea of combining multiple agents with different personalities and different skills to research and debate information in order to provide the best answers is not new. They're not the first one who are doing it. We actually covered how you can do this on your own with the regular ChatGPT on episode 167 that is called Build an AI Dream Team that Works for You in Different Roles and Personalities. So you can go and check out episode 167 if you want to build a team of super powered agents that will work for you and will help you make better decisions.

12:53But doing this in an agentic approach with much more data and much more capabilities is really appealing. And at least based on the benchmarks, Grok4Heavy is taking the best advantage out of this concept moving forward. Now, it sounds expensive. It's$300 a month compared to, let's say, most of the tools that are$20 a month. But if you can solve your biggest problems in your business, in your company, in your research, in your distribution, in whatever aspect of the business that you want, and it's going to cost you$300 a month, it's a no-brainer, right? You may have access to more or less the best consultant on the planet for$300 a month when most consultants charge more than that per hour.

13:32So if you look at it from that perspective, $300 a month is nothing. Now, XAI also is releasing Grok through the API. So you'll be able to have access to it through a paid subscription to Grok or through the API in order to develop whatever you want and integrate with it to get the best results out of what you are building. And they're working to integrating it into all the hyperscalers, such as Amazon, Google, and Microsoft Azure. And they also announced their future releases under the Grok for Umbrella or under Grok in general, which is an AI coding model that is coming out in August. So next month, a multi-model agent development platform in September, a video generation model in October.

14:10And Elon also announced that he plans a game lab that will encourage AI powered game development. And they expect to deliver that this year. And he's expecting the first full-fledged high-end AAA game to be built with Grok probably next year. So that means think about a top game like FIFA or Microsoft Flight Sim or GTA games at this level developed through vibe coding with Grok that gives you an understanding of how sophisticated and capable he believes this platform will become within the next 12 months. Now, it is not going to be the first time that Elon Musk makes claims as far as timelines that are completely unreasonable, but let's say he's wrong by 100 % it's going to take two years, that is still very significant and profound in its implications of software development, because these are games that currently are developed by hundreds of people over a couple of years.

15:01Being able to do that through vibe coding is nothing short of incredible. And it also means other software that's sophisticated will be able to be created by single individuals over a relatively short amount of time without necessarily having great experience with coding and developing software. But it is very rare that you're going to get news related to Elon Musk without some kind of controversy. And so in this particular case, very shortly after its release, Grok was briefly restricted after posting anti-Semitic remarks, including using different Jewish last names and associating them with negative association, as well as praising Hitler.

15:37and so the model was restricted for a short while and they removed part of its system message that was requesting it to use politically incorrect answers in order to make it more interesting. So this has always been part of Grok is making it a little unhinged as Elon Musk called it and making it more fun and lively and less vanilla than all the other models that obviously comes out with a price and Elon summed it up very shortly. He said, Grok was too compliant with user prompts, too eager to please and be manipulated essentially. that is being addressed. So they took it down, they made a small change, and hopefully now it's not going to do these kind of things.

16:13But I'm not sure that's the last we're going to hear from X doing weird things. That's just the way they build the model. They build it to be a little edgy, and I think they cannot care less about the model doing these kind of things. And if there's going to be a backlash about this topic or the other, then they're going to address it one by one, versus going through six months of rigorous checking of what the model is capable of. You take it to the extreme of what people can use this model to do if it's better than PhDs on every single topic on the planet, like creating different mass destruction weapons or any other negative thing you can think of.

16:42And now you have a bunch of PhDs in your pocket without necessarily being tested whether you can use them to do bad things. I think that is really, really scary. And again, I'll be really surprised if the level of scrutiny on what the model can and cannot do before releasing it is something X cares about, at least as deeply as some of the other labs. Now, Elon went further and even said that Grok4 is a bold alternative to established players emphasizing greater transparency and fewer content restrictions. So on one hand, that is a good thing. You don't want to necessarily have OpenAI or Gemini or anybody else decide what you're going to see and what you're not going to see.

17:16But the flip side is, if X wants to become a real alternative to these other platforms and they want enterprise users to use X as a real alternative, it has to be controlled. You cannot have a system running your business, basically acting on your behalf for customer service, etc., saying things that you cannot say as a company. whether they're anti-Semitic or other negative tone, because then it will get your entire company in trouble. It risks your clients and you can't have that. And so I think while Elon's personality is definitely reflected in the way X runs and the way Grok works, they cannot keep on doing this if they want to turn this into a real alternative on the enterprise level.

17:54And I'm sure that they do because that's where the big money is. And to give you an example, how you may get a backlash from that, on July 9th, Turkish court banned X Grok from being available in Turkey. It's because Grok was issuing insults to the president, Erdogan, his wife, his late mother, and Kamal Ataturk, who is the founder of modern Turkey. So Grok wasn't just doing this against Jews, it was just being Grok and being very loud and profound about everything because that's how they built the model. And now they got completely banned from Turkey. The X team obviously immediately apologized and addressed the situation, but this is a one of fixes for a much bigger problem.

18:31And I don't think it's a problem. I think it's just their approach. That's the way they built the model. And so these kinds of things are going to keep on happening with X. This time they were banned by an entire country, maybe not a huge deal for X. But let's say that the model now is integrated into a large enterprise and it shares insider trader information with the public because it thinks it's going to be helpful. And then you get in trouble with the SEC. And then there's a lot of backlash that could happen from things like that. So this is just one example of how this could go terribly wrong for a company.

18:59And so if I have to guess, they would at least build an enterprise level that will be more tamed and more controllable, and maybe the public version will stay what it is today. But that's not the end of the news from X.AI. It almost confirmed on Twitter or on X that X.AI's Grok AI, including the latest version, which is Grok 4 that I just released, will serve as the voice and the cognitive core of Tesla's Optimus robot. So they're going to take their most advanced AI from X.AI and combine it with Tesla robots. They're also going to embed it into Tesla cars. So you'll be able to have a conversation with your car.

19:30I'm not sure exactly what it will enable. I'm sure it's going to be an extra fee per month to use it. But I assume if you're going to be paying for Grok, you'll be able to get it in your Tesla as well. I personally drive a Tesla and I'm very curious to see how is that going to work and exactly what it will enable me to do with the car. I'll be willing to pay the fee just to test it out and see what benefits it provides beyond the capability to have a conversation, browse the web and get information. But it shows something very obvious. it shows that Elon Musk as a person and X as a company and Tesla as a company has a very serious level of trust in the capabilities of Grok 4.

20:03Otherwise, they would not put it in the cars and they would not put it in the robots. It means that from a maturity perspective and from a benefits perspective, it makes business sense. Now, again, Elon is known for making big bets and taking risk in his businesses, but he's also known for being able to make the technology that he develops extremely successful. And this convergence of what Elon is doing right now is fascinating and can have incredible impact on the society of the world. So if you think about, we just mentioned three of his companies, X that was Twitter, XAI that is developing Rock, and Tesla that is developing cars and robots that are self-driving and the robots will probably have a lot of autonomy.

20:41Combine that with two additional companies that are Neuralink and Starlink. Starlink that can deliver potentially AI capabilities through the web to any location on the planet and Neuralink in the long run, the ability to connect that to a human brain. And you see that the complete vision, the complete integration of everything that Elon is working on is way beyond science fiction. But I have zero doubt that that's where Elon is pushing it. Think about having a chip in your brain that can access AI capabilities anywhere on the planet through Starlink and connect and work together with robots and cars and other machines and potentially the streets and everything around it is going to have sensors and AI embedded into it as one unified environment.

21:21This sounds like the Borg from Star Trek, I know, but I truly believe that this is where Elon is going, at least in his vision. Is that even possible? How long will it take? I don't know, but I have a feeling that we'll get to find out in the next decade or maybe And from a XAI company perspective, the announcement of Grok 4 came on the same day as the announcement of the departure of XAI CEO. So Linda Iaccarino stepped in as the CEO of XAI in seriously troubled times for the company. It was shortly after Elon bought Twitter for$44 billion. Iaccarino was the former NBC Universal ad executive, and she joined X on May of 2023, so just over two years ago.

22:02and she was the one that was able to more or less stabilize the platform and get the company from very serious term oil to a relatively steady company. Now, she claims that through her 10 years, she was able to grow X by 12%. That being said, if you look at similar web data, it shows a drop from 388 million users to 311 million. That's a very significant decline during that time. But she was able to stabilize the spiral that was happening at that time where many, many publishers were leaving the platform. She also obviously had to deal with Elon as the face of the company, which is not easy because he would say whatever he wants, regardless of what is going to be the impact.

22:40And if you remember in an interview when he was asked about publishers leaving X and losing ad revenue, he said, and I'm quoting, go fuck yourself when he was referring to the publishers. And so you combine that with the need to run a successful company that is built on getting publishers to give you money. I'm sure Yaccarino's job wasn't easy. Well, either way, she's on the way out. They have not announced a new CEO yet. I assume that it's not easy working for Elon for a very long time. And especially now after the merger with X.ai that he is the CEO of, and now she's technically running a company that reports to another company that Elon Musk is running.

23:16That doesn't make it easier for sure. Where does that leave us from an X.ai perspective? Well, where it leaves us is with a company that when it got started was the joke of the industry and couldn't make a logical sentence connect properly to now the leading company from a benchmark perspective. I said that when they announced Grok in the beginning, I said, I'm not going to bet against Elon, especially with the amount of data he's going to have from Tesla, from Starlink and from X, formerly Twitter. And so he has the data, he has the distribution. He built Colossal, which is the most capable AI computer data center in the world today.