227 | AI is now outperforming top human experts in coding & real-world tasks 🤯 Sam Altman's favorite ChatGPT feature just dropped, $100Bs pouring into AI infra, and the economy may never be the same — key AI news for the week ending on September 26, 2025

26:34So they are shorthanded all the time. A lot of people are leaving because it's not an easy job to do and people are looking for other things. And so this new robotic approach, the idea is to replace the employees that they're lacking instead of the employees that they have right now. And their initial focus for these robots is data collection for ship construction and classification. This means that ABS standards on how to do data collection, data analysis, and so on are going to change in order to make them a better fit for these robotics. So the way the shipyard work is being handled is going to change dramatically.

27:10So it's not just, oh, let's look for efficiencies, is let's find different, more innovative ways to do this, because now we're going to have skills that were not available before. I shared with you a few months back when a big infrastructure project was happening in my community, and there were dozens of people walking around the streets with metal detectors and spray cans, spraying where different lines are running. I'm like, there is no way robots are not going to take this over and probably can take this over right now of just collecting the data of where the different lines are passing. So when you're digging, you're not running into them.

27:38This is just one very simple example that I saw with my own eyes. But these kinds of things are going to get replaced by robots probably faster than we can even think. Now, if you think about a shipyard environment, and I obviously never worked at a shipyard, but I can definitely imagine how it looks like, it's a hectic and dynamic environment that requires adaptability and humanoid robots with quote-unquote thinking brains is the perfect solution in order to be able to fit into the existing operation without making significant infrastructure changes. Now, speaking of robot and changes that are coming, Chinese startup Ahead Forum has unveiled a humanoid robot head that mimics human emotions.

28:14And it's really weird to look at it. So they released two different videos of their robots. One looks like a human, the other one looks like an elf. They call it the elf series, so it makes sense, that just has incredibly realistic facial expressions. And it is blinking and move its head and lips and everything in a way that is nothing but weird and scary to me. But what they're saying is even weirder and scarier. They are claiming that, yes, they can't anticipate exactly the full trajectory on the development of this and the deployment of this, but they anticipate that within no more than 20 years, robots will be able to look exactly like people and we won't be able to tell the difference.

28:53This is the mother of all science fiction, right? If you think about a lot of movies that we've seen where we can differentiate between the robots and the humans, they're claiming this is coming within less than 20 years. Now, if we think about any other technological prediction that happened in the past that always fell short, that might be 15 years or 10 years where we're going to have robots that are indistinguishable from actual humans. I don't like this thought at all, but the technologies are being developed right now. And with the right business model, this is where it is going to go. So now after talking about agents and robots and how they're going to impact the workforce, let's talk about why I think this is not stopping.

29:29And this is the crazy investments that are happening right now in the infrastructure that is supposed to drive all of this. OpenAI just announced that as part of their Stargate initiative, they are going to be building a capacity of seven gigawatts of new data center in the near future. So they're building five new sites plus an expansion of their Abilene, Texas and CoreWeave projects, and they're doing it way ahead of their initial schedule. So we shared with you their crazy partnership and deal with Oracle for$300 billion worth of extra capacity, 4.5 gigawatts of extra capacity to be specific across several different locations.

30:10And as I mentioned, part of it is just the expansion of their existing cloud setup in Abilene, Texas, which could be housing 400 ,000 GPUs by mid-2026, so less than a year from today. Now, a quick recap on what is Stargate and how it was announced. So it was launched in January at the White House with President Trump, and it's a partnership between Oracle, SoftBank, and OpenAI in order to create significantly more compute for OpenAI to keep them competitive, presumably against China. So if you remember, Ted Cruz said, we should have a very light touch regulatory approach because we are in a race with China.

30:46Basically tells you this current administration will do everything required in order to keep the US companies ahead, which means allowing them to build, consume a huge amount of power, and generate significantly more environmental impact than probably any project in human history. But now we have a new deal from this week where NVIDIA is going to invest$100 billion in OpenAI via a phased approach that will generate 10 gigawatts of GPU-powered data centers owned by OpenAI. So the goal here is that NVIDIA is going to provide these data centers or these GPUs to the data centers that OpenAI is going to own versus the compute that they rent right now from Microsoft and Oracle and so on.

31:27So how is this still going to work? NVIDIA is going to fund 4.5 million new GPUs for 10 gigawatt data centers starting in the second half of 2026. Just to put things in perspective, the deal with Oracle is for 4.5 gigawatts. So they're talking about 10, that more than doubles the amount of compute just from this crazy large infrastructure that they are planning. Now, there are a lot of discussions about the bubble and how crazy these investments are and so on. There are a lot of people that are saying, oh, so the way this is going to work is that NVIDIA is going to invest$100 billion in OpenAI, so they have$100 billion to actually pay Oracle out of the$300 billion that they need to pay Oracle that they don't have.

32:07And this way, Oracle will be able to make a lot more money that will be able to invest in NVIDIA and doing the circular generation of money out of thin air with billions and billions of dollars. The reality is I don't think this is completely accurate because I think this, for several different reasons, one, this is presumably on top of the previous investment. So these$100 billion in NVIDIA new chips is supposed to be new processing capability, not as part of the previous project. And the other is it was made very, very clear by NVIDIA that these are going to be owned by OpenAI and not by Oracle and or somebody else.

32:41So while I understand the concerns and while I understand the fears of the crazy bubble that is potentially shaping in front of us, I don't think this is exactly the right place to say, oh, here's a smoking gun of what's actually happening. As another data point on how crazy is the need for AI compute is right now, AI chipmaker Cerebrus, which makes the world's fastest inference chips right now. So putting things in perspective, the GPUs that NVIDIA is generating that has grown them to be the most valuable company in the world is used right now both for training AI models as well as for inference.

33:19Inference is when the AI is generating outputs. It's called inference. And so the fastest inference chip right now is by Cerebrus. There's another company called Grok, Grok with a Q, not Grok with a K, to be separated from XAI's Grok large language model. So Cerebrus is currently raising$1 billion at a$8 billion valuation in a move that is showing you very, very clearly what is the demand for what they're doing. There's definitely a future in which GPUs are going to be used more for training new models and less for inference because these inference chips are significantly cheaper and run significantly faster, providing higher efficiency at the inference time.

33:58And if you think about it, inference is going to be the core of compute in the future as we will need to train less models because the existing models will just going to be good enough, not necessarily the existing today, but the ones that will exist in the future are going to be good enough for most things. And then just using them is going to be the need. And for that, there are better chips than the NVIDIA GPUs, at least right now. Now, in addition to this crazy investment by OpenAI on new compute to deliver everything that they need to deliver, whether it's training models or creating inference or building devices and so on, OpenAI is also committed to investing$100 billion in backup servers.

34:34So this is stacked on top of the$350 billion of rental of compute that they have between now and the end of 2030, combined with the additional$100 billion from NVIDIA. you kind of get the point. These are numbers we've never, ever, ever seen in any industry, in any infrastructure of anything ever before. But this is what OpenAI are committed to as of right now. But they are not alone. So Microsoft just announced that they are converting the Mount Pleasant, Wisconsin, Foxconn 8th Wonder of the World building into a powerhouse AI data center. Putting things in perspective, they're taking an existing building that was built to do something else and failed and turning it into a high capacity data center with hundreds of thousands of NVIDIA GB200 GPUs and fiber connections across everything to make it extremely fast.

35:26They're claiming, by the way, that the amount of fiber optics there is 180 ,000 kilometers, which is four and a half times, which means it can go around Earth four and a half times. The good news is the cooling for this facility is all closed system liquid cooled, which is presumably cooling 90 % of the space without wasting any new water, which is good news from an environmental perspective. However, the energy that this will consume is 250 megawatts solar farm matched by fossil fuel usage and other renewables. To put things in perspective, a few green organizations from Wisconsin are saying this equals to over 4.3 million homes.

36:10And to put that into perspective, Wisconsin as a whole has 2.8 million homes. So this one facility will require more electrical power than one and a half times the entire homes of the state. Now, here's another data point for you for how much investment is going into this AI boom. A new report that was published on CNBC talks about the seven elite private tech companies that has completely sky rocketed in the past two years to a combined value of 1.3 trillion. These are not publicly traded companies. These are private companies. So obviously the biggest one is OpenAI with a valuation somewhere between 350 billion to 500 billion.

36:52Elon Musk's ex-AI with a$200 billion valuation, who by the way, just raised another$10 billion just shortly after the previous$10 billion raise earlier this year. And the valuation went up from 150 billion to 200 billion in just a few months. Anthropic with$178 billion valuation, which is kind of weird when you think about their being behind XAI when it comes to valuation, but Elon is Elon and he can do stuff that a lot of other people cannot do when it comes to valuation of companies. Databricks after that with$100 billion in valuation and three additional non-AI companies such as SpaceX, Stripe, and Unreal.

37:30Now to explain how much investment these companies has racked up, AI startups have vacuumed$65 billion across 19 companies just this year. That is 77 % of all private capital. I have to say this again, 19 companies has gotten$65 billion, which is 77 % of all private capital this year. This tells you the crazy amount of concentration of investment that's going into the more promising AI companies out there. That being said, the CEO of Forge Investment, Kelly Rodriguez, says, we've not seen this in any private market ever. Companies that are growing 100%, 200%, 300 % on numbers that are already pretty big.

38:16And she doesn't only just mean the valuation, she means the revenue as well. Let's just think about Anthropic. Anthropic went from $1 billion in revenue at the end of last year to$5 billion right now, not even a year. But with all that money, most of it going to infrastructure, Deutsche Bank has sound the alarm as far as the really scary underlying economy problems that this is hiding. So their head of research, George Saravellos, has mentioned that AI CapEx has outpaced all consumer spending to drive the US GDP growth this year. But in order for the tech cycle to continue contributing to GDP growth, capital investment needs to remain parabolic.

38:56This is highly unlikely. Or basically what he's saying, he's saying that if you take out the crazy investments that happened in AI in 2025 in the US, and you take that out of the current equation, the US economy would have been in a recession. So while we're seeing the economy growing and we see GDP growing, what he's basically saying that is fueled by AI and AI infrastructure, and the rest of the economy is actually shrinking right now, which is definitely not good news, unless the output of the AI will generate significant actual growth in other industries that will then compensate for that really big jump in investment, specifically in AI infrastructure.

39:37In a Bain & Company report that was released this past week, they project that AI needs to generate$2 trillion in annual revenue by 2030 in order to just fund the 200 gigawatts of global compute that it is committing to right now. And even with all the savings that they're seeing right now, they see a shortfall of$800 billion in the equation as of right now. Do I agree or disagree? It's very hard for me to say. These people are way more knowledgeable than me when it comes to doing these kind of calculations. I will just say that I think AI will drive such an incredible shift in everything we know about most industries and most markets that we can't project how it will actually impact the economy.

40:19because we're walking into uncharted territory, but it's so far from chartered territory that is very, very hard to project, even for people who are highly experienced in doing these kind of projections. There are also optimistic groups like Goldman Sachs that think that AI will eventually create a significant productivity across multiple markets. But there are also other negative views that are saying that the S &P 500 right now, as an example, has a huge exposure to AI. And if the AI boom collapses, anybody who's an investor in the S &P 500 is going to get a very serious hit. So you hear voices on both sides of that equation.

40:51I think both of them are guessing and speculating based on a lot of unknowns. And we'll just have to play this out and see how this works. So that's it for the deep dives. What we've learned from them is that there is the greatest investment in history in AI infrastructure right now, more than any investment in anything else in the past, in a really, really high speed, at a really high scale, fueled by a global race to get first to AGI, ASI and beyond, and that the companies that are behind it are investing amounts of money that sound completely legendary. However, each and every one of these things is leading to results that potentially put at risk every single job or many, many jobs across the known economy without anybody knowing what other jobs, if other jobs, will emerge to replace them.

41:39So overall, this doesn't look too promising to human jobs in the at least near future until we figure out what other jobs we can do in the AI era. Now let's switch to rapid fire items. There's a lot of stuff to cover. The first segment in the rapid fire is going to be about new releases and new capabilities in existing models. The most interesting one this week is that XAI announced Grok4Fast. And Grok4Fast is one of those mini models, like we've seen before with O3 Mini and so on, that are almost as good as the existing models. So Grok4Fast performs almost as good as Grok4 across multiple benchmarks, but it comes at a 98 % price cut.

42:18So you're going to pay 2 % to get almost the same exact outcome. And in addition, this new model comes with a massive 2 million tokens cost context window, which is the top of what is being offered right now, which is at the high end of the scale, similar to Gemini 2.5 Pro and way above ChatGPT as an example. It is also very good at multi-modal capabilities, and it has improved dramatically in code generation and basically everything else you want to use AI for. And in addition, based on third-party analysis, it is currently by far the frontier runner when it comes to cost-efficient intelligence.

42:53If you want to get solid intelligence at a very low price, Grok4 Fast is your number one option. And I'm sure a lot of people that are developing against APIs are going to switch to it, at least until the next thing comes over. So right now you can get a top-of-the-line capability for$0.20 for 1 million input tokens and 50 cents for 1 million output tokens, which is way more competitive than any other platform out there right now. Now, speaking of compute and investments and how much more of it will become available and what will enable and being able to run models in a more effective way and so on, leads us to the next announcement.

43:29Sam Altman, like he likes to do, dropped a Cryptic X post this week that talks about a new compute intensive offering. So I will read the exact quote from there, over the next few weeks, we are launching some new compute-intensive offerings. Because of the associated cost, some features will initially only be available to pro subscribers, and some new products will have additional fees. Our intention remains to drive the cost of intelligence down as aggressively as we can and make our services widely available, and we are confident we will get there over time. But we also want to learn what's possible when we throw a lot of compute at today's model's cost at interesting new ideas.

44:11What does that mean? Well, nobody knows. There are a lot of speculations, which leads us to the next topic, which is several different people have been reporting that they're seeing a new selection of models in the OpenAI drop-down menu and in their API that is called Alpha Models. Under these Alpha Models, two separate agentic models appeared. One is called Agent with Truncation, and the other one is called Agent with Prompt Expansion. These are not very good names, but A, there's a very good chance that was not supposed to be exposed to the public, and B, OpenAI has not been very good at making up names in general, so that wouldn't surprise me if that ends up being the file name of these new features.

44:48When using these features, what it was activating is agent mode, but in a very specific setup that will still allow it to run browser and use tools and so on, but with more capabilities than it had before. Now, shortly after these reports came out, these models were rolled back and were not available anymore. But is this what Sam Altman was referring to? New agentic capabilities running within the chat GPT agent mode? Nobody knows, but it feels like this is the direction that they're going, which is deploying more agentic capabilities that by definition will require a lot more compute. But as Sam mentioned, we will know in the next few weeks.

45:24And I promise to keep you posted as things are happening. But there is a feature that OpenAI did roll out this week, which is a feature that Sam Altman calls, and I'm quoting, Today we are launching my favorite feature of ChatGPT so far, called Pulse. This also points to what I believe is the future of ChatGPT, a shift from being a reactive to being significantly proactive and extremely personalized. So what is this feature and how can you get it? So first of all, it has been deployed first and foremost to the pro subscribers, the$200 a month people, and only on the mobile application. It is generating between five and 10 personalized cards as you open ChatGPT every single day.

46:03And these include an image and a little bit of text. And when you click on it, it takes you to a more detailed view of the same topic. And these are going to be personalized topics that you care about. This could be soccer updates or travel plans for a specific destination or a diet-friendly the menu and so on and so forth. And you click on those and it will take you to that specific information. The other thing is it is capped by the number of these cards that you're going to see. At the end of it, there's a statement saying, great, that's it for today. And the idea is intentionally to move away from the endless scroll time waste of social media.

46:37So basically, it's to give you high value, organized and highly personalized information that you need to start your day. This could be personal, this could be business related. Now, this feature integrates with their existing connectors, such as Gmail and calendar parsing, which means it can surface emails and agendas and so on. And it is connected to the chat's long-term memory for context and understanding of what you're working on, what you like, and so on, which will become more and more personalized over. In the demo that was provided by Christina Wadsworth Kaplan, she showed how Pulse Auto added London running routes for her based on her jogging history and also offered pescatarian tweaks to her dinner reservations in order to help her find the right restaurants.

47:21OpenAI's new CEO of applications, Fiji Simo, blogged, we're building AI that lets us take the level of support that only the wealthiest have been able to afford and make it available to everyone over time. And Chachapiti Pulse is the first step in that direction, starting with pro users today, but with a goal of rolling out this intelligence to all. But I want to go back to Sam's tweet. And I'm going to read the second half of it again, and then I'm going to refer to it. This also points to what I believe is the future of ChatGPT, a shift from being all reactive to being a significant, proactive, and extremely personalized.

47:58What does that mean? It means we get a glimpse into the future, an AI that is not a chat, but that actually is proactive and participates in our day-to-day activities across the entire day from personalized to business things, fully connected to the entire universe of our personal knowledge, again, both personal and work. Now, if you combine this with wearable devices, such as glasses or whatever it is that OpenAI is developing, and that means we're going to have devices that are always aware, always collecting information about what we do. And you understand that these tools will become extremely powerful.

48:37It will know everything about us and it can become the most incredible support sidekick that we ever could dream about. It can help promote healthy habits. So don't eat this or eat that or go work out more frequently or get up and walk a little bit or connect with friends and so on. But it can also do better time management that we're doing right now. Allow us to focus when we need to focus. Allow us to plan things that we're currently finding hard or don't have time to plan as deeply as we want to do and assist with personalization and prioritization of things and so on and so forth across basically everything that we're doing.

49:13This is really exciting and really, really scary at the same time. A, because the complete invasion of privacy of technology into everything in our lives, which is just going to be an extension of what we know today from, as an example, Google knowing everything about me because I have all my data in the Google universe and I use an Android phone. So it's just going to be a very significant expansion of that. But in addition, it is scary because I feel that we will become completely dependent on this technology. If this technology will be assisting us with making every decision along the day, how will we know to make our own decisions without this technology?

49:48And I think very quickly we'll become addicted to this while ChatGPT is saying this is going to be different than social media because you won't be able to scroll forever. I think the negative impact might be much more significant than social media because it will be integrated into literally everything that we're doing. Now, from OpenAI to Anthropic in a topic that's actually related to OpenAI. And Anthropic has got a lot of bad press in the last month or so for things not working well with Anthropic code generation capabilities, which was the most advanced and most capable code generation tool so far.

50:20It has been the default for most developers around the world, and it has fueled the crazy growth of Anthropic from $1 billion in revenue to$5 billion in revenue in less than a year. Well, there's been a lot of speculation around what's actually happening with the Claude models. A lot of people were speculating that Anthropic is throttling down the models in order to save money and so on and so forth. Well, Anthropic just came out with a report that is sharing that they found three significant bugs that has together led to the outcome that Claude responses were suboptimal. I'm not going to dive into the exact technicalities of these bugs, but the reality is the three bugs combined could have affected up to 30 % of cloud code users at the end of August through the beginning of September when they started deploying solutions for that.

51:10That took them about two weeks to deploy, so through mid-September, they were able to roll out the entire solution for all these different issues, which presumably is supposed to solve the problem. They swore they're not throttling the models and that they're providing the best capabilities that they can to everybody all the time. The reality is that the big winner out of this is OpenAI. And the reason for that is as people were getting very unhappy with the degraded results from Cloud, GPT-5 has been shining as the most capable code generation tool right now. And I assume, and again, there's no statistics that I found, but I assume that a lot of people have jumped ship from running Anthropic behind the scenes on the development processes to running GPT-5, which leads me to one of the most interesting new phenomenons of the AI era, which is how easy it is to switch from one platform to the other.

51:59So if you think historically, if a critical infrastructure of your company were not happy about it and you wanted to switch, let's say you wanted to switch from SAP to Oracle or the other way around, it would have required months of planning, more months of execution, and millions of dollars in investments to switch from one core infrastructure to the other. Right now, switching from Claude to Chachapiti to X and vice versa as the underlying infrastructure in your tech stack takes seconds and practically has no cost. Like in your IDE, as an example, your code generation environment, you can go in the back end, click the drop down menu and select a different model and be up and running.

52:37And that's it. No cost, no investment, no infrastructure change. And in this reality, it needs to be very scary to the companies that create this infrastructure because until they have very significant integrations into companies' systems, they will be able to be replaced with a simple drop-down menu and with no cost to the companies. So there's significantly less stickiness than we are used to in the tech industry. Another big interesting release this week comes from ByteDance from China. They just released Sea Dream 4, which is the latest image generation model that is supposed to be as good as Gemini Nano Banana, and it is supposed to be competing directly with that.

53:17That's why they released this model. It is very good at generating images. It is very good at generating text. It is very good at editing images, and it runs 10x faster than their previous model. So a lot to look for. I haven't tested it myself, but it comes at a very attractive price point of$30 for a thousand generations at a very high speed and at a higher resolution than Gemini Nano Banana. It is currently ranking second on the LM Arena text to image ranking, but a very, very close second. So Nano Banana has 1 ,152 points and Seadream 4 has 1 ,151 points. So this is a fraction of a percent difference in the models, and it is cheaper to run Seedream 4.2.

54:04Now, if you're just generating it on your regular chat, it doesn't really matter. But if you want to use it through the API, then it matters a lot because it's going to be significantly cheaper to generate very similar results. It is also ranking second, but with a bigger spread on image editing capabilities with Gemini Nano Banana with 1337 points and Seedream with 1313. Again, still a relatively small spread, but not as close as the other one. And they are all far ahead of all the other models behind them. Another interesting feature that was released this week was released by Meta. Meta just launched Vibes on September 25th, which is a dynamic feed of AI generated short videos that invites users to browse, create, and remix content and music and styles to generate new content out of that.

54:49So you can either create videos from scratch or remix existing content from all the other platforms and reshare them back to the other platforms. So if you see a Meta AI video on Instagram, as an example, you can tap to bring it into the Meta AI app, remix it, add music, change the style, and repost it back to Instagram, Facebook, etc. So this is a whole new interactive environment for generating, editing, and remixing existing AI-generated videos. Users have generated over 10 ,000 videos in just the first 24 hours of this feature. Do I think it's going to take over our feed? Probably not. Do I think it's going to play role in the future feed?

55:34100 % because you'll be able to create really cool videos that were just not possible before unless you were Disney and were willing to invest millions of dollars in generating something cool when now anybody can generate them. And so I have zero doubt that they will be a part of every feed that we see in the future, but I do think people still want to know what other people are doing, what people are achieving, and so on and so forth. So I don't know how big this will turn. Time will tell. And let's switch gears from talking about features and releases to really big strategic partnerships that have rolled out this week.

56:06The first one is Luxshare, which is a key supplier to Apple. That's their shares surged 10 % after there's been reports of a partnership with OpenAI on their next AI-driven device. So this Chinese manufacturer that has been assembling the Apple AirPods and the Vision Pro is working on a prototype device together with ChatGPT, which is leading to a lot of speculations whether that's the device they're developing with Johnny Ive or not. It is not completely clear, but what is clear is that OpenAI is going all in on devices. The information is shared this week that in addition to the crazy investment in Johnny Ive and his team, just as a quick reminder, that was$6.5 billion early this year, OpenAI has been recruiting aggressively hardware engineer and designers directly from Apple.

56:53So while there's been 10 previous Apple engineers in OpenAI in 2024, they've recruited over two dozens in 2025. And in addition to Lux's share, OpenAI has approached other suppliers of Apple devices for potential components, such as components to a speaker model in one of their future devices. The latest rumors is that the first OpenAI device that is targeting a release in late 26 or early 27 is going to be a speaker without a screen. But either way, it puts a lot of pressure on Apple, not just because OpenAI has been stealing some of their top engineers, but because of what it means to Apple.

57:34Apple is a device company. They're a hardware company. They've built an incredible brand that millions and millions of people in the world are dependent on. But they've been failing in the AI race time and time again in the last few years. And with the soaring popularity of ChatGPT as a tool that everybody knows as AI, and with the soaring and with the growing popularity of wearable devices, definitely the glasses from Meta. But again, once OpenAI comes up with a ChatGPT-based device, X number of millions of people are going to buy it. And eventually, these will start taking market share from iPhones and other Apple devices.

58:09And I'm certain Apple is aware of it. The window of opportunity is closing because if OpenAI really releases this product within about a year from now, this puts a very, very serious pressure on Apple to do something and do something successful and do it fast, which they were not able to do in the past two years. Another huge partnership this week that actually connects to one of the earlier topics that we discussed in this episode, is OpenAI and Databricks has forged a multi-year$100 million deal to enable a new agentic development platform based on OpenAI tools inside the Databricks environment.

58:46So the goal of the partnership is to enable enterprises to easily develop AI agents using the OpenAI capabilities while connecting it to their proprietary data that is hosted on the Databricks platform. This obviously makes perfect sense to both companies, and it will provide a lot of value to the companies that are using the Databricks platform. Another big announcement this week when it comes to partnership, the General Services Administration, also known as GSA, has signed a deal with XAI for 18 months agreement to deliver Grok's latest models to anybody in the government for$0.42 per organization in the government.

59:23In addition to delivering the models for practically free. XAI is committing engineers to help agencies implement Grok rapidly and successfully. Plus, they're suggesting training for the employees in order to explain to them how to best use the models in a good way. We reported about similar deals that were signed with Microsoft and OpenAI just in the past few months. But my biggest question really goes back to training, right? So providing the models to the government is great. The fact that the companies are willing to do it for free. I don't think they're doing it out of generosity. I think they just understand that after the free segment is done, they will be able to make billions of dollars from government employees being used to using their models and having it fully integrated into multiple different processes within the government.

1:00:07The government being the largest employer in the US makes it all makes perfect sense. Also, it's good relationships with the administration. So it all makes great sense. The problem is training. Who is going to provide the training to the government employees in order to use and leverage these tools effectively, that still stays unclear. What I see in many companies that I work with is like, oh yeah, we're investing a lot of AI. We got everybody licenses for whatever, whether it's JGBT or Microsoft Copilot and so on, but they're not seeing any results. And in many cases, they're seeing negative impact because there's no proper training to the employees.

1:00:39And I have a feeling that a similar thing on a much larger scale is going to happen in the US government. So I really hope that the government will figure out how to provide adequate and continuous training to government employees on how to leverage AI effectively and safely so we can all benefit from this of potentially getting better government services cheaper and faster. From government, let's switch to legal for a minute. XAI just filed a lawsuit on September 25th accusing OpenAI of stealing trade secrets via hired away employees. Now, this is not the first and probably not the last lawsuit between Elon Musk and Sam Altman or between XAI and OpenAI.

1:01:17It is very clear that this battle is not going to stop. In this particular case, as I mentioned, Musk is claiming that Sam is stealing talent, but not just the talent, but they are actually using trade secrets that were developed inside of XAI as they're getting the employees into OpenAI. The lawsuit names three former XAI employees, two engineers and one senior executive that allegedly passed along proprietary source code and other business secrets as they were joining OpenAI. Now, I'm not going to dive into the whole history between these two individuals, but this thing doesn't seem to quiet down.

1:01:52It's just getting worse and worse. I don't know if to say that's getting ridiculous is the right word to use, but it is getting ridiculous. They're acting like five-year-olds and mostly Elon is acting like a five-year-old. The other thing that this brought me to think about is that engineers are jumping ship in this race all the time from one company to the other. And if this somehow makes it to court and somehow makes it to trial and somehow gets to a situation where the court is saying, no, you cannot do this or you have to do that, then this is going to stop something that's been happening for decades and definitely happening in the last two years, which is people that are moving from one company to the competition.

1:02:28It is very, very hard to draw the line in the sand of what they know versus what is considered stealing trade secrets. And it's going to be very interesting to see how this evolves. Again, And I don't think this is going to go very far. I definitely think this is just a personal battle between Elon and Sam. And this is just another way for Elon to put another stick in the wheels of OpenAI. Again, I don't personally think this is going to go very far. Another big partnership that we actually announced a couple of weeks ago, but is now actually available to the public, is Microsoft 365 Copilot now includes Cloud Sonnet 4 and Cloud Sonnet Opus 4.1 as part of its selection of models.

1:03:04There's even a try-clawed button in the researcher app inside of Copilot, and it allows the users to switch to clawed Opus 4.1 for deep reasoning tasks, replacing the chat GPT models. One of the several different reports have showed that Anthropic models significantly outperform OpenAI when it comes to integration with Excel and PowerPoint. So that's another reason for them to be well-integrated into the environment. And it makes perfect sense to Anthropic because it provides them access to huge enterprise audience. And it makes perfect sense to Microsoft because it allows them to diversify and reduce their dependency on OpenAI models.

1:03:44I would not be surprised if other companies are rolled into there. If XAI is going to really be continuing at the trend they are right now, they're probably going to be next. That's it for this weekend. We'll be back on Tuesday with another how-to episode. In this particular case, we're going to show you how to create incredible videos that can promote your brand across social media and other channels in seconds without teams and people and influencers and so on, all on your own while using AI with all the pros and cons that comes with doing what I just said. If you have been enjoying this podcast and finding value in it, please hit the subscribe button and please share it with other people that can benefit from it.

1:04:22Literally just click on share right now and just think of a few people that can benefit from listening to this podcast and just send it to them. I'm sure they will thank you. I will thank you as well. And for now, keep on exploring AI. Keep sharing what you learn with me and with other people wherever you can. It will help all of us to reduce the potential negative outcomes out of AI and have an amazing rest of your weekend.