322 | AI Data Centers Are Now a Bipartisan Punching Bag, Claude Tops User Satisfaction, MHS is the New MCP for Hardware, and More Important AI News for the Week Ending August 28, 2026

27:36And they have maybe a little more ability to do that, but maybe they don't have to. We're going to talk about the opinions about what people think about the different models right now. So we'll see that Anthropic has reasons to be able to charge a premium price, but we'll get to that in the later segment. But for now, what I want to talk to you about is the financial aspect of it. So OpenAI are burning through cash. Anthropic, on the other hand, are potentially breaking even depending on the month and how much they're spending on CapEx and other things right now, but Google has$300 billion of advertising business that they can use to potentially discount their models and subsidize them to be competitive on the low end in ways that OpenAI and Anthropic just cannot.

28:18So while it seems that right now that Google is not competing at the frontier, they're definitely playing a stronger and bigger role on open source models with Gemma and on their cheaper model with GPT 3.7. So Google, while they're not competing at the frontier right now, they're definitely competing on the low end. And with Gemini 3.7 Flash, they have a very solid contender on the low end. We just had a really interesting conversation in the AI Weekly Friday Hangouts. That is a community meetup that you're all invited to join. It happens every Friday at 1 p.m. An amazing group of people that just care about AI and want to learn from one another.

28:56And one of the people mentioned that he started using Gemini 3.7 Flash for coding, and he's extremely satisfied. And it is now his favorite model to develop software. And he has been using Anthropic and OpenAI top models so far. So he has a solid benchmark. So again, on the low-end models, Google are playing as a serious contender in that game. And maybe that's their focus, but they also have much, much deeper pockets and they can afford to subsidize their models in order to stay in the competition. Now, the fact that AI is getting cheaper is making it more available in more places and for more use cases.

29:34And this is starting to have a noticeable impact on other markets. The biggest example came this week from Baidu. So Baidu is the Chinese Google, right? It's a huge company with really large revenue and their revenue for Q2 was 4.62 billion, which missed analyst expectations, and it is down 4 % year over year. This is the fifth consecutive quarter of decline when their online marketing services is down 19 % year over year to 13.1 billion yuan. Now, in their statement, their real estate slump in China and weak consumer demand has been cited as macro causes for that. But the reality, AI is disrupting their search-based advertising business.

30:23It is not happening to Google yet. Google is actually seeing the best quarters they've ever seen when it comes to advertising revenue, but it is definitely taking a toll on Baidu. And I assume it will have a similar impact on Google as AI usage growth as a way to replace traditional search. Now, what is saving quote-unquote Baidu for now is that their core AI powered business segment, basically providing infrastructure for AI to run, has gone up 25 % year over year and is now half of the general business revenue. GPU cloud revenue jumped year over year. AI cloud infrastructure is up 50 % year over year.

31:03What does that do to the overall financials of the company? It is a very different ballgame selling advertising versus selling cloud services. So their net income attributable to Baidu specifically is down 68 % year over year. This is obviously a very big difference in how profitable the company is. And because they're now making huge investment in CapEx, their free cash flow has swung to negative and it's now 1.1 in the negative compared to positive cash flow previously before this big expand and before this big decline in revenue. This news obviously sent their stock down, their US listed shares down 14 % in the beginning, closing 20 % lower at a$90.8 a share, where companies like Morgan Stanley is now setting their price target at 80.

31:53So that's still much lower than it is after this big decline. Doesn't look very good for Baidu right now. And it makes perfect sense. Again, if you are able to generate extremely high margins based on marketing revenue, this is one thing. If you have to fight all the big giants while investing hugely in infrastructure in order to sell it and maintain it, and you're significantly less profitable, that is a very different ballgame and playing a big role in why people are using less search and hence less ad revenue for Baidu. The other thing that this is doing, this fight to get cheaper models to do stuff for you is showing up in acquisitions and M &A as far as where the real value is.

32:34So more and more companies and more and more people around the world are understanding that you don't have to be married to OpenAI or Anthropic or anybody else. You can shift around different models as needed if you have the right infrastructure and if you have access to open source models and closed source models and you can shift at will, this is extremely valuable. This is leading to something we talked about last week when OpenRouter is going to get acquired by Stripe for something between$7 and$8 billion. OpenRouter allows companies to switch between one model and the other very easily. I've been using them for multiple different things for the last couple of years, and so are many other people.

33:13They are currently processing over 10 trillion tokens every single day across 400 models and more. Another relevant, similar acquisition that is fueled by the same thing, there are rumors that Hugging Face are now potentially going to get sold at a$13 billion or more price tag up from$4.5 billion valuation in 2023. So huge jump in their valuation, the potential acquirer, NVIDIA. Well, NVIDIA seems to be acquiring everything or their names seem to be tied to potential acquisitions of more or less anything, but it actually makes a lot of sense for NVIDIA to purchase them. NVIDIA has an endless amount of cash right now, and it's growing every single quarter, and owning the platform that hosts most of the world's open source models makes perfect sense for them to be able to have a more complete ecosystem.

34:03So what does this all mean? First of all, it means that people are definitely paying attention to how much they're paying for AI right now. I'm working with multiple businesses, large, medium, and small, and I can tell you it is a big deal in all of them. And in large enterprises or midsize companies, it is maybe the second most important conversation around AI after data security. And so that has a very clear impact on what we're seeing. As I mentioned at the beginning of this article, a 43 % decline in inference pricing over 10 weeks in US enterprises. That is very significant. And it is going to continue and it is going to continue for two reasons.

34:42Reason number one, the companies will create cheaper models to use because of the competition. Reason number two is companies understand they don't need frontier models in order to do the day-to-day work. In tests that I've done recently, I tested all the different models on day-to-day stuff. And I can tell you that I was able to do very helpful work with Haiku from Anthropic and Luna from ChatGPT and similar open source models and get very decent results while paying significantly less. Before that, I didn't test. I just assumed I have to use Opus for some things or at least Sonnet for some things.

35:16And what I've learned, that's not actually the case. And the more people test these things and the more people understand which use cases require what level of model, we'll allow companies to use lower level models to achieve day-to-day tasks, which will put the pressure even more on the labs to discount prices because otherwise they will run out of business. So what does that mean to you? First of all, it is good news. We are going to get better intelligence at a cheaper price. The second is go test for yourself your own use cases across several different models and see which is the cheapest one that will give you consistent results on the tasks that you are performing and stay there until the next cycle of models come out and then test those and start all over again.

35:58Now, the last story that I want to spend more time on before we go into rapid fire, I'm not going to dive too deep into this one because it was just announced, but I just think it is going to change everything again. So on August 27, Anthropic launched what they call Model Hardware Standard. This is, if you want, the hardware equivalent of an MCP. This is a universal interface that allows AI agents to directly operate lab equipment, such as manufacturing machinery, robotic arms, microscopes, etc. So they're following the same, if you want, playbook they did when they released MCP at the end of 2024.

36:35They're releasing it as an open source standard, which hopefully will become the standard across all the different companies around the world in a similar way that MCP has took off. MCP right now has a million plus monthly downloads. Just shows you how wildly used MCP became. Anybody who is a little more advanced in AI is using multiple MCPs every single day, myself included. Now, this new standard, MHS, again, model hardware standard, has already been deployed with some deployment partners such as Denentech, Carnegie Mellon University, HHMI, Janlia, and Quera Computing. And they also have vendor hardware partners with AWS, Danaher, University Robots, and a few others.

37:22All these companies are currently either building support for MHS into their products or testing it in deployment of using actual devices. Now, the interesting thing is the results that these initial partners are describing. Carnegie Mellon University in a serial dilution experiments have stated that running experiments is now three times faster than previous methods because they're using MHS. Genentech have used it in BCA protein essays. Genentech have used it in internal processes in which they actually found some issues with the initial testing, including some critical failure that they've surfaced that has been since fixed, but they are using it in their manufacturing capabilities.

38:03Quera Computing, which is a quantum laser system, achieved 99.3 autonomous laser lock recovery, which is the highest autonomy figure disclosed by any early adopters. I, Genelia Research Campus has unified seven vendor programs that have no unified integration between them into one orchestrated rig using MHS. So think about taking seven incompatible vendor stack and replacing them with one unified interface without going through months or potentially years of integrations just by connecting everything to this new protocol and standard. It is completely changing the way things can happen in labs and in manufacturing facilities.

38:46And these come from very different angles, right? So these companies span from genomics to protein science to quantum computing and university research, all finding very useful use cases and achieving interesting results by using this new integration. What is that telling us? It is telling us that Anthropic is focused way beyond just software, right? They now want to integrate their intelligence into hardware as well. They're also into hardware in many other ways. We discussed that they have recently hired a hardware executive, Caitlin Kalinowski, that was previously at OpenAI and Meta and Apple.

39:20They are assembling a custom cheap team. So they're planning to build their own silicon. So they're going to be the same as all the other labs and have their own computer chips built to their needs. So they're going to play way beyond just the model creation. They're planning to integrate their models into A, their own hardware that they're building, as well as into more or less any hardware in the world through this new protocol that they just introduced. Now, the bigger question obviously is security and safety. That should be a big concern here, especially as these models are getting significantly more sophisticated and capable of doing really catastrophic things.

39:58So we talked about many times about the stopping of the development at OpenAI or slowing the development at OpenAI. We talked about the Glasswing project where Anthropic released their Mythos models only to a short list of companies to enable them to find different loopholes in systems. The problem while this is happening, and these models are currently being used to evaluate the different vulnerabilities in software, nobody can fix them fast enough. So based on the recent information from Glasswing project, vulnerability that it found is potentially over 23 ,000 vulnerabilities across over a thousand project.

40:34They're finding these vulnerabilities at a rate that nobody can fix fast enough, even with AI coding. Think about now taking this to the physical world. Think about allowing AI agents to control physical machines and devices in labs and manufacturing facilities. And think about the risks, both cyber risks as well as physical risks in this case, that this may introduce, not to mention the sci-fi story of now AI controlling manufacturing, so they can manufacture whatever version that they need of whatever device that they need, so they can build more of whatever they need. So if you go down the Terminator path, this is definitely starting to sound like Cyberdyne, at least to me, and I'm definitely a sci-fi geek, and I really like the Terminator movies, so I may not be the right benchmark for that.

41:22But allowing the AI to control all the manufacturing machines, because it can do it significantly faster and better, also comes with at least theoretical baggage in my mind. But that being said, this protocol now exists. It's out there. And if it's going to catch at the same speed as NCP, we're going to see more and more companies around the world allow the machines to manage, control, calibrate, and do whatever they define to actual hardware using this new protocol. And as the AI becomes better and better, the abilities to do that will become more and more important to companies to implement because that is going to make them significantly more competitive, which will put pressure on them to actually do that, which will open the door to a lot of really bad things potentially happening.

42:03I'm not saying they will happen, but potentially happening. So that's it for the deep dive this week. Now let's switch to rapid fire. I'm going to go really quickly. There's a lot of really interesting things that happened. And since we talked about Anthropic, let's stay with Anthropic and Claude. They made a lot of interesting announcement this week. So the first one, as I mentioned earlier, there has been a satisfaction ranking survey that has been released, and Claude leads user satisfaction, outperforming Gemini and ChatGPT in the YouGov survey. That survey has been open from March 1st through July 31st, so running for a while, and ended very recently.

42:38And Claude holds the highest AI satisfaction among the current and former users, and it's achieving a net score of 56.2, significantly ahead of Google Gemini and OpenAI ChatGPT, while Apple Siri and Elon Musk Grok lag far behind. So putting the actual numbers in place, Anthropic with Claude at 56.2, Gemini with 46.5, ChatGPT at 46.0, so pretty close together. Then Perplexity with 43.7, Alexa 42.6, Copilot with 37.2, DeepSeek with 35.1, and Apple Intelligence with 32.4. And really far behind are Grok with 24.6 and Siri 23.7. So what does that mean? Not much from a technical perspective or implementation perspective, just from a public opinion perspective.

43:27There's obviously a very strong swing towards Anthropic and Claude since the beginning of this year. When I started using Claude heavily in Q4 of last year. I spoke to people and they never heard of Claude before. And now almost everybody I talk to are like, yeah, yeah, we like Claude. We switched to using Claude from using ChatGPT. So from a public opinion perspective, there's definitely been a very strong swing in the past year towards liking Anthropic and liking OpenAI a little less. I must admit, I'm now on the swing back. There's more and more things I'm doing with Codex and other ChatGPT capabilities.

43:59To be fair, in parallel and in conjunction and working on the same project in collaboration between Claude and Chut. Another big announcement from Anthropic this week is that they're unifying the memory of the regular Claude with Claude Cowork, which is going to provide Claude more knowledge and more context, which means it will be able to support you even better. So previously, the memory from the regular Claude Conversations was not available to Claude Cowork, and now Claude Cowork can actually tap into the memory of regular Claude Conversations, which is going to provide a lot of value to people who are regular Claude users and now are switching to using Cloud Cowork.

44:34As somebody who's been using mostly Cloud Cowork and Cloud Code for the past year, that is not a big difference to me because I built my own memory and data and context systems and layers that actually connects to all my files and all the systems and my entire ecosystem, basically. Not a big deal for me, but if you've been a regular Cloud user and now you're going to transfer to using Cloud Cowork, that is a big deal for you. They've also made updates to the real-time memory updates and management and a lot of other memory-related stuff that is going to provide value to people as Claude learns how they do and what they do and so on.

45:08Anthropic made another big announcement this week, which Claude now launches its own built-in browser in its desktop app. So it does not need now to go to your Chrome extension and control your actual Chrome window. It can actually run a browser within the Claude environment. The biggest benefit is that it's completely isolated from your own browser. So all the passwords and links and things that are saved in your browser stay in your browser and Cloud operates in its own browser environment inside of Cloud where it does not have access to your universe. It does also mean that it doesn't necessarily need the extension for Chrome, Edge, Firefox, and so on.

45:44So that's a new development and a new release from Cloud similar to how it works in ChatGPT recently. Anthropic also released a few new capabilities that existed in a limited test so far and is now available to the public. This unlocks AI agents with general availability for computer use, Skills API, and Files API. So as of August 20th, Anthropic Computer Use Tool, the new browser tool that I just mentioned, Skills API and Files API are now generally available on the cloud platform, which enables and opens the door for a lot of capabilities that previously were just for specific selected companies.

46:22An interesting aspect of this new computer use tool, it is now eligible for HIPAA-regulated workflows under Anthropic Business Associate Agreement, which opens the doors to a very wide range of companies and use cases that were previously banned from using AI for this thing, which is showing you that these tools are becoming better and better at keeping the information safe if highly regulated controls are allowing this process to happen. And from Anthropic to OpenAI, the first big news I already hinted about earlier, and that is OpenAI data center head departs as part of a really big shakeup.

46:59More and more people are leaving OpenAI. And as I mentioned, this is strange, especially as they're heading into their IPO. I don't know how much of it is intended, meaning they're switching people to get to the IPO with a better setup and a better team, or people are leaving for whatever other reasons. But it definitely doesn't look good from the outside. More than a dozen executives have left the company from the beginning of this year. And it doesn't stop at the executive level. Business Insider is saying that 13 to 14 other senior leaders have left across various divisions in the company as well.

47:32So a big departure on the leadership team, which again, doesn't look great, but specifically the biggest loss this week is Chris Malone. He's joined in March of 2025, OpenAI's data center strategy. So critical to their current and future growth. Now, this is a big deal because Malone was one of the key figures behind OpenAI's Stargate project, which is a$500 billion effort to build data centers in collaboration with some really big names, including Oracle, NVIDIA, and SoftBank. So how will that impact OpenAI? I don't know, but it is definitely another big name that is leaving the company in the recent months.

48:13Another big movement of a big name this week is Luke Metz. Luke Metz is joining the superintelligence team in Meta, but he has a very interesting history. He was a key researcher in OpenAI. He left OpenAI to join Mira Moratti's Thinking Machine Labs in 2024, rejoined OpenAI in early 2026, and now is transitioning to Meta. So top scientist that started his journey at Google brain. I don't know if it's a money thing or an interest thing, probably a combination of both, but another big name in the AI space that is jumping ship in recent months. Staying on OpenAI news that you need to know, OpenAI just reinstated their five-hour daily limit for ChatGPT plus codecs and work features.

49:00That limit existed before they removed it, and they're bringing it back effective on August 20th. Why are they doing this? The formal reason is that it allows OpenAI to, quote unquote, smooth the compute load on their system. But the reality is it's combined to what we talked about earlier in this episode. They're giving big discounts. This discount allows people to use more tokens and OpenAI has to somehow balance the amount of tokens they're using because they don't have unlimited capacity. And the way to do this is to stop people after five hours and not let them use the cheap tokens that they now have access to.

49:34So you're paying less to get the tokens, but you can't use more tokens because you're going to be capped. So So OpenAI looks cheaper, and at the same time, they can still control the amount of compute that they need to have in order to allow this cheaper AI to be available to people. Now, from news from specific companies to a few interesting pieces of news specifically about agents. The first one is in a study that was published in Science Advances, revealed that over a thousand AI agents, including advanced models from OpenAI, Anthropic, and Meta, spontaneously aligned with a majority opinion, achieving 100 % consensus, even without explicit instructions or memory of past interactions.

50:15Now, the problem with that is that researchers are warning that this is a way of agents to create, if you want, a power of the majority that can lead to catastrophic failures in critical multi-agent systems. So if you have built a really sophisticated process that has multiple agents that are supposed to balance one another, it is eventually, potentially will fail because all the agents will end up agreeing on whatever topic instead of balancing each other out, which is obviously not what you're planning if you're building them to balance themselves out. Staying on the topic of agents and use, Cloudflare just unveiled Kitesurf, which is a lean browser engine that is built specifically for AI agents.

50:57So what does that mean? It means this browser is significantly more efficient for agents to use compared to Chromium-based browsers. If you want specific numbers, their resource efficiency dropped to approximately 3.1 to 3.8 times less CPU and 4.7 to seven times less memory than running the same exact agents on a Chromium browser. The idea is that this browser was architected and designed specifically for tasks such as screenshot and HTML extraction that will support agents in what they need to do. I talked about this in this podcast many times before, that the web, as we know it today, will change.

51:37We will have increasingly more and more agents, unless humans actually visiting websites and crawling the web, meaning the way we address the world, the way we build our websites needs to change in order to support agents better. And now it's not just the websites, it's also the actual browser that they're going to use is going to be built for agents and not for humans, which means again, your website over time will have to change in order to work better in that browser versus Chrome or other browsers that exist today that are built for humans. Staying on agents and combining it with talks about OpenAI.

52:11OpenAI are hosting the WebMCP Challenge. It is a 10-day competition, and they're inviting developers to build applications that leverages the WebMCP protocol, which is a standard that they just recently announced, and it is supposed to help AI agents interact directly with websites through structured tools. Now, the prizes are not that high. The top winners can win$3 ,000 in cash, but they're also going to win one year of ChatGPT Pro, a Codex Micro keyboard, and additional prizes that are going to be built. But the big deal is you will be able to get your name out there as somebody who won an OpenAI competition.

52:49Now, what is WebMCP? As I mentioned, it's something OpenAI just announced in partnership with other companies, and it is an experimental open standard that allows websites to expose structured tools specifically for AI agents to be able to use them directly, which allows AI agents to complete tasks faster in the browser. Same exact concept as we talked about with the new browser from Cloudflare. Now, this new initiative by OpenAI is starting as another one is ending. OpenAI just finished what they call Build Week Hackathon, which drew over 47 ,000 builders from 186 countries who submitted over 8 ,000 projects across these eight days.

53:27The goal is to demonstrating how AI tools can enable non-programmers and domain experts to create functional software. As somebody who has been doing this more or less every single day for about a year, I can tell you it is life-changing. You can literally build any kind of software you need for yourself, for your family, for your company, and so on. The thing that people don't tell you is that software needs to be maintained and supported and need specific data security features and access points for people and agents and so on. And that makes it significantly more complicated than just building the software itself, but the ability to build software is definitely there.

54:04Now, staying on agents, an interesting announcement from NVIDIA this week. They have just announced that their Agentic Variation Operators, or AVO for short, is a new architecture, and using this architecture, they achieved a 100 % score on the ARK AGI 3 Interactive Reasoning Benchmark, completing all 83 levels across 25 different environments. And what it is showing this breakthrough that by combining autonomous agents' performance through a system-level design allows it to have longer memory and better overall outputs across long and complex tasks. Now, to put things in perspective, Claude Opus 5, the second most strongest model from Anthropic, achieves 30 % on Arc AGI 3 under a standard evaluation.

54:55So achieving 100 % tells you how well built the orchestration platform is to enable the agents to run across it and achieve all the different goals successfully. This is the world we're going into, where single systems that we all think are extremely capable but still fail on basic tasks, like all the models that we know, will be able to overcome this by building agents that will operate jointly on these tasks and will be able to achieve probably everything that we can achieve that can be done on a computer. And now to three interesting things that companies have announced this week, some of them through partnerships.

55:32Salesforce and Anthropic has launched what they're calling CloudForce. It is an integration that is designed to embed Cloud directly into the Salesforce platform for sales professionals. What it has is 37 pre-built skills specifically for salespeople that can do multiple tasks like email composition and record updates directly inside of Salesforce using the Cloud AI environment. Salesforce shares rose 12 % following the announcement. Now, to put things in perspective and obviously connected to Mark Benioff, which is Salesforce colorful CEO, Mark Benioff said that, and I'm quoting the nonsense of the saspocalypse.

56:15I think it is time for it to stop. But the reality is it is not stopping. If anything, I think this proves exactly the other way around. And the other thing that makes it funny is the fact that the shares rose 12 % after they are down 22 % this year and 20 % in 2025. So there is a huge decline in the last two years in the stock of Salesforce because more and more people understand they don't necessarily need it and they can deal with AI stuff that will be significantly more efficient and user-friendly than Salesforce. And now the fact that I can actually do this with Anthropic basically says I don't actually need Salesforce.

56:52I need Salesforce data structure. So then I can use AI on top of that. So I think while this drove the shares up 12%, in the long run, I think it will lead to exactly the opposite results. And another huge interesting partnership this week is between SpaceX and NVIDIA. So these two companies, these two giant companies are teaming together to launch a Vera Rubin NL72 system space optimized CPU designed for a genetic AI to orbit in Q4 of next year. All these talks from Elon Musk about putting AI data centers in space is going to be at least an experiment within just a year and a quarter from now.

57:34So NVIDIA is going to build a new version of their chips that are going to be optimized for being used in space. It is positioned as the first CPU built for agent, and it's designed to accelerate orchestration, code execution, and data processing, which is the infrastructure of the new generation of the agentic era. Now, how much can Elon actually launch to space to make it viable and interesting? Nobody really knows. The numbers he's talking about right now just don't add up. There's just not enough launch capacity in the world, and there won't be enough launch capacity in the world to make this worthwhile to the numbers that he's talking about.

58:10But even if he's off by an order of magnitude, it will still be the largest data center in the world, and it is going to be in space. So the vision from Elon Musk, again, like in every other thing that Elon Musk ever touched, is slowly becoming a reality. And this is definitely a huge step in the right direction, partnering with NVIDIA on this process. And the last piece of news that I have to end on, and I couldn't skip despite the fact it probably does nothing for you in the immediate future, but it has huge implications in the long run, is if you somehow missed it, the humanoid robot Olympics happened in China this week.

58:45And there were a lot of really crazy, incredible things in there. Now, the biggest news out of that was a Chinese humanoid robot that completed a 100 meter sprint in 8.86 seconds. How crazy is that? The fastest human to ever complete the race in a record that will probably never be broken is obviously Usain Bolt from Jamaica, who run the distance at 9.58 seconds in 2009. The first robot to beat that actually did this on Saturday at 9.39 and then immediately after that, a new robot broke the record with 8.86. Now to tell you how big the event was these robots participated in multiple other sports as well, achieved incredible success, and it had over 1 ,300 competitions involving more than 2 ,000 robots competing in 51 events, including running, long jump, table tennis, ballroom dancing, and including some real-world scenarios like household tasks, hotel services, and emergency responses.

59:46So while this is somewhat an entertainment event, the meaning or the implications of that are endless. It means that robots that are not crazy expensive can now do things that were very clearly human physical skills that at least I thought will take a while for robots to compete with. And they're already doing that. Now, putting things in perspective, most of the participants are from China by a very big spread. By the way, the latest stats is from a deployment perspective. The Chinese currently control over 95 % of actual deployed humanoid robots in the world. So all the news we're hearing about all the other companies, the current pace, China has a complete win victory over every other company in the world when it comes to actually deploying robots that do work.

1:00:35But developing these robots to do these specific sports capabilities is showing really advanced mechanical capabilities that will be translated into robots doing other things. So if you think about how much technology we have today in cars that started with race cars, you understand that this is how the same thing is going to happen here. You develop things for the most demanding environments. You learn from that all the time, and then you can build it at scale, simplified to service every need that you have. I really think that the next big wave that is going to impact blue collar jobs will come from robots.

1:01:13And I think it's going to happen faster than we actually anticipate. That's it for this week. We'll be back on Tuesday with a fascinating how-to episode. Don't forget to share this podcast with people that you think can benefit from it. It will take you literally just a few seconds. Click on the share button on your phone, unless you're driving, and then think about three to five people that can really benefit from knowing more about AI and just send them the link to the podcast. I'm sure they will be grateful. I will be grateful. You'll be doing the right thing. I will be feeling good about yourself.

1:01:43And while you're at it, it will be nice if you can rate and review the podcast on Apple podcast or Spotify. And don't forget to check out the multi-agent orchestration course. The next cohort that is now open is in November. So if you want to learn how to build really sophisticated automations that can support your personal life and your business, and you still want to do it this year, this is the time to sign up. But that's really it for today. Have an amazing rest of your weekend, and I'll be back on Tuesday.