October 12, 2026
2026-10-12
https://www.youtube.com/watch?v=5_TdyFZ0JNM

[0:10] We go live in ten.
[0:24] Every morning, the future moves faster.
[0:29] The models, the money, the power, and the people building what comes next.
[1:08] From Austin, Texas to the world.
[1:15] AI News, Real Context.
[1:18] This is the Austin Wilson Show.
[1:21] I'm Austin Wilson, your human in the loop.
[1:29] Good morning, and welcome to the Austin Wilson Show where we're breaking the AI headlines that matter to you today.
[1:36] My name is Austin Wilson, and today is Monday, 10/12/2026.
[1:42] We're coming to you live from the AI capital of the world, Austin, Texas.
[1:46] And we've got a fun show today.
[1:49] A lot of really important things to cover, and I hope you guys enjoy it.
[1:54] Today, uh, we're going to talk about the OpenAI cofounder and his view on the AI gap.
[2:02] Anthropic is quietly adding a Bitcoin miner.
[2:07] Nvidia may buy an open model lab.
[2:10] Satya Nadella, the Microsoft CEO, says AI needs an emergency break.
[2:17] Apple hires the AI podcast team, and Robot Labs are filming you at work.
[2:24] Let's get into it.
[2:27] First up, we are talking about, uh, somebody that is really important in the AI community.
[2:33] One of the OpenAI founders posted that the gap in who gets AI is getting wider, meaning who understands AI, who's using it the most.
[2:45] Uh, this is Andrej Karpathy.
[2:47] Um, he is one of the OpenAI founders.
[2:50] He's actually now at Anthropic and is one of the leading minds in, uh, AI today.
[2:57] He posted on x, uh, over the weekend.
[3:01] He said, dusting off this older tweets from April be...
[3:04] April because this gap in shared understanding of LLM capability is widening.
[3:10] It's now less, uh, two groups of people speaking past each other and more a sharp funnel.
[3:17] He breaks down the the math on this, kind of his his napkin math around AI usage worldwide.
[3:24] He said around 6,000,000,000 people, about 75% of the population, have barely come in contact with LLMs at all.
[3:32] Around one to 2,000,000,000, about 20% are casual and infrequent users of a free tier, you know, ChatGPT like like product.
[3:42] Uh, he said this group sees, you know, derpy chatbots and treats them a bit like a a better Google search, uh, a writing aid, or etcetera.
[3:51] Maybe an agent tries to book you a flight.
[3:53] He said, I have nontech friends who reasonably, in my opinion, say they have not much use for it at all.
[4:01] Even many professionals outside of math and code are in this tier.
[4:04] For example, execs and many other functions spend a lot of their time talking to other people.
[4:10] So while they understand what is happening intellectually, it is still secondhand and a bit abstract.
[4:17] Now we get to professional use of frontier grade LLMs in math and code.
[4:22] Somewhere around 20,000,000 people, about point 2% of the population see firsthand that large complex projects that used to take them weeks or months can now be completed by agents with a prompt, uh, building apps, copying apps, translating apps, decompiling apps from binaries.
[4:44] Uh, this has all happened very quickly and recently.
[4:46] Less than one year ago, Karpathy says, I was writing code manually by hand, typing memorized computer code commands into a code editor character by character, occasionally pressing tab to auto complete a little chunk of code.
[5:01] And then he gets into his his last part.
[5:03] And finally, we get to the about 5,000 people, about point 00006% with access to frontier grade systems internally.
[5:14] Uh, these are folks at the AI labs that have even more advanced models, uh, than the general public.
[5:21] The external world has seen the preview.
[5:24] Um, it looks like swarms of thousands of agents collaborating over weeks on software mega projects, minting zero days, running cyber attacks and defenses at machine speeds, discovering new science, advancing the frontier of mathematics, things that would have taken top professionals in the industry years of work.
[5:42] Meanwhile, human review and comprehension are starting to fall behind.
[5:46] For example, people are still involved in the, quote, unquote, archaeology of the OpenAI Hugging Face incident from many months ago.
[5:53] Mathematicians may be pouring over the 722 manuscripts on frontier mathematics for a while.
[6:00] The funnel is driven by a combination of factors.
[6:03] One, the impact scales with ambition, problem size, and horizon.
[6:07] A question with a paragraph answers, uh, barely stresses the system.
[6:12] You need a reservoir of big difficult problems that you really care about.
[6:16] This aspect drives the consumer professional dimension of the funnel.
[6:20] Two, the jackness of the system, which I've written about a lot separate...
[6:25] Separately, capability peaks in domains that are digital, verifiable, and economically valuable.
[6:31] This is because LLM capability emerges from reinforcement learning on verifiable rewards on a curated environment mixture driven by revenue potential.
[6:41] This aspect primarily drives the area, e g math and code dimension of the funnel.
[6:48] Three access, free tier, paid tier, internal.
[6:50] So this is the weirdness of the moment.
[6:52] The general public has mostly not interacted with these systems.
[6:56] When they have, it looks like a derpy chatbot.
[6:59] The majority of professionals still see only a modest uplift.
[7:03] A small sliver of professionals are experiencing the vertigo of the curve going vertical, and it's all happening at the same time.
[7:15] What does this mean?
[7:16] That was that was a lot to cover there.
[7:19] And so the the reason I read that full tweet is it, um, fully encapsulates what we're seeing as a society, uh, as it relates to AI and superintelligence and advanced intelligence and technology, uh, that we haven't come across before.
[7:41] I I think this is a new paradigm where instead of...
[7:46] Typically, on a new technology, you have an adoption curve where early adopters get into it first, and then it kind of expands from there.
[7:54] But the rate of expansion, uh, and capability of the models is going at such a pace.
[8:01] They're they're getting so intelligent so fast that it's actually kind of having an inverse relationship from an adoption curve perspective where we see the, uh, the adoption has caught on, but the technology, uh, is moving so fast that that adoption cannot keep up with it from a kind of a general perspective.
[8:22] And this is really important to understand in that, and I think it it piggybacks on something that we've covered in the past and that there's a gap between understanding the the reality and the capability of what these models are are able to do.
[8:43] And this creates, uh, this it it creates a lot of not great emotion and, uh, just understanding around it.
[8:53] What do I mean by that?
[8:54] I mean that the general public, uh, is is mostly uncertain, uninformed, uh, maybe fearful of what the capability of these AI models are.
[9:09] Right?
[9:09] And so what does that mean?
[9:10] It it it means that we're we're looking into a future where AI and technology is capable of driving things forward at such a rate that humankind is is really, you know, struggles to keep up with.
[9:24] Right?
[9:25] But it means that there's a future where...
[9:27] And and, you know, the upside on it is so good.
[9:30] Right?
[9:30] The upside on it is amazing.
[9:32] That's what's exciting about this technology and these models is that the the positives coming out of it are phenomenal.
[9:39] Right?
[9:40] Right now when we don't understand them, we...
[9:42] It's easy to just, uh, focus on the negatives and forget what kind of, uh, we're doing all of this for.
[9:50] And so what I wanna do is just encourage you guys.
[9:53] Get out there.
[9:54] Get into the models.
[9:55] Read about it.
[9:56] You know, using ChatGPT and Claude as much as you can.
[9:59] Ask Claude, uh, you know, what's the best way for me to use AI in my day to day work?
[10:06] What's the best way for me to leverage this technology, uh, to maximize the positive benefits in my life today?
[10:14] And I believe that if you get into the technology, you're going to see, uh, positive effects that that are gonna reverberate, you know, through your life and help you, uh, achieve things that you never thought were were possible.
[10:27] And that's where it's really exciting is when we get these tools to society in general, uh, we create an opportunity where, uh, everybody has super intelligence at their fingerprints, uh, fingertips and can achieve some amazing things.
[10:41] So I wanna start today with that message.
[10:43] Get out there.
[10:44] Let's close that gap and ensure that everybody has the capability to use these fantastic tools.
[10:52] Alright.
[10:53] Uh, Anthropic has reportedly signed an unannounced multiyear cloud deal with Iron, a former Bitcoin miner to get Claude more compute.
[11:02] The information says analysts estimate one two hundred and forty five megawatt site could bring in about 1,300,000,000 a month once it's running.
[11:12] Anthropic is pushing partners to build faster, leaning on ex crypto miners like Hut eight, Terra Wolf, and nScale.
[11:24] Compute is the current bottleneck.
[11:26] Every month, a site switches on early is worth real revenue to Anthropic and, uh, AI providers.
[11:35] Analysts say a 245 megawatt site could add $1,300,000,000 plus in revenue per month.
[11:42] The information sourcing on the iREM deal, iREM already leases capacity to Microsoft under a 2025 contract.
[11:50] Uh, no dollar value, term, or megawatts disclosed for iREM, and the $1,300,000,000 figure is an analyst estimate, uh, not confirmed yet.
[12:01] Now as we get into it, we continue to have this conversation.
[12:04] We continue to see this, uh, week over week and sometimes on a daily basis, these new compute deals, which means, again, you know, the underlying hardware that the AI models are running on top of.
[12:18] And it doesn't seem like, uh, Anthropic or OpenAI or any of these AI, uh, frontier companies are pulling back or or slowing down or seeing a reduction in their their need and capacity to run compute.
[12:36] And so, again, additional deals coming up here where, in this case, it's kind of interesting.
[12:43] The, uh, the previous use case for this data center was, uh, crypto mining.
[12:51] Right?
[12:51] And now we're seeing that convert to AI inference and and compute running for AI models.
[12:59] Uh, this is a standard thing that, interestingly enough, just slight backstory, is that crypto miners and and crypto mining setups run on, uh, or or can power very efficiently AI workloads.
[13:14] And so, uh, it translates really nicely, and and, uh, what a what a great deal for here.
[13:20] Uh, it sounds like Anthropic is gonna continue expand their compute capacity and looking for more and more ways to run these models.
[13:29] So I would not be surprised if we continue to see deals like this as it doesn't seem like their, uh, capacity to gobble up compute will be satiated anytime soon.
[13:42] Moving into the next story.
[13:44] Nvidia is reportedly in talks to buy Reflection AI or at least deepen its stake days after the startup shipped its first open model.
[13:54] The Financial Times says one option is an acquihire hiring staff and licensing tech, which could dodge a long antitrust review.
[14:04] Nvidia has already put in about $800,000,000.
[14:08] A deal could land within weeks.
[14:11] And, we send...
[14:12] You see Nvidia moving into the model, uh, level of the stack.
[14:19] Right?
[14:19] So Nvidia would own a US open weight lab built to compete with cheap Chinese models.
[14:26] Nvidia could take Reflections team and technology without a full acquisition.
[14:32] Financial Times sources, uh, Reflection launched Beam its first open weight model last Monday, and sought a $25,000,000,000 pre money valuation in April.
[14:43] Neither company commented, and Aqua hire deals are drawing closer scrutiny from the FTC.
[14:51] Alright.
[14:51] We love to see it again.
[14:53] Uh, Nvidia coming in to get get into the different layers of the stack.
[14:58] Right?
[14:58] So we've talked about this before, but the stack, you know, basically comprises of the infrastructure layer where we have the hardware, the compute, everything, all the machines, the data centers running these models.
[15:11] Then we have kind of the middle layer, which is the model layer, right, where we have the AI models running and developing smarter and smarter intelligence.
[15:21] And then the the top layer is the application layer.
[15:24] Uh, this is where we've got, uh, you know, different applications like Sierra or Elise AI or, uh, different AI tools.
[15:31] We talked about Cal AI last week that are applications specific to a certain use case that are running on the model.
[15:39] The model is running on the hardware that compute.
[15:42] Right?
[15:42] So interesting in the the, you know, previous story.
[15:45] Anthropic is looking for more compute.
[15:47] It's at the bottom of the the stack.
[15:49] Right?
[15:50] Then we've got, uh, Nvidia, uh, is at the bottom of the stack.
[15:55] They provide the chips for the data centers to run compute on, And we continually see Nvidia develop and grow up the stack.
[16:04] Right?
[16:05] And so they are big proponents of open source models, which means, uh, anybody can use these models.
[16:11] They call them open weight models, which means that you could download it and run it on any hardware that you want.
[16:18] You could own that piece of software, uh, essentially for free.
[16:23] Running it might cost some money, but you you can own that software for free.
[16:26] It's freely available.
[16:28] And Nvidia has their own open source model.
[16:31] We saw last week, uh, Reflection, uh, AI released their, uh, latest open source model, which is competitive with the, you know, Frontier Chinese models from an open source standpoint.
[16:43] And so Nvidia continually trying to position and compete.
[16:48] They are the big dog out there, you know, $5,000,000,000,000 thereabouts valuation and the, I believe, most valuable country or or company in the world right now.
[16:59] And so Nvidia is trying to secure their position and maintain a a strategic, uh, um, a strategic spot within the stack.
[17:10] Right?
[17:10] So in in...
[17:11] Interesting to see this.
[17:12] Of course, Nvidia is already an investor.
[17:14] They've already put $800,000,000 into the into the company.
[17:17] Now will we start to see pushback from the FTC around acquihire or or mergers or or acquisitions that are like this where we start to see maybe some, uh, consolidation and anticompetitive stuff.
[17:33] I think it's way too early to to say that anybody, any one company is creating a, you know, monopoly or an unfair advantage here, uh, but Nvidia is definitely trying to be strategic and work in that.
[17:48] You know, curious to hear from the audience.
[17:51] Should...
[17:52] If you have a perspective on this, do you think Nvidia should own the models that run on the same chips that they're selling?
[17:59] Interesting to see, or should they remain separate, stick with the chip game, and let others focus on the models?
[18:07] Coming out of Microsoft CEO, Satya Nadella says every AI model needs an emergency break that a human can pull mid task.
[18:16] In a Saturday post on x, he said companies should assume a model is compromised and contain it from the start, and then log every meaningful action as a tamper proof human readable evidence.
[18:30] This is interesting.
[18:31] And Microsoft sells models to enterprises.
[18:34] Its CEO now argues none should be trusted by default.
[18:39] Uh, separate the model from its harness and externalize the controls.
[18:44] Nadella Nadella is making this argument that that it follows alongside an Anthropic disclosure around agents that they could be filing false police tips and 20 visa applications that they should not have been doing.
[19:01] Uh, he named no Microsoft products specifically, no timeline, no enforcement mechanism, uh, behind the idea.
[19:11] This is, uh, really interesting coming out from ChatGPT seeing the...
[19:16] And and this went around x all weekend, a a lot of conversation around this idea.
[19:22] And this comes back to an idea that we spoke about last week in that, currently, the the pace of development is moving so fast and the agents are, uh, move at a pace that humans can't keep up with.
[19:37] We, you know, we talk about the, um, you know, solutions to math problems, reference the OpenAI Hugging Face incident earlier this summer, and the fact that humans are still trying to unravel some of these, uh, advancements or or or events that ever...
[19:56] Have occurred or or that the AI has committed or completed.
[20:01] You know?
[20:01] So so we're...
[20:03] Humans are trying to understand and unravel these things.
[20:06] The pace at which we can do so is much slower than the artificial intelligence or superintelligence is moving forward, if that makes sense.
[20:16] So intelligence is, you know, going a 100 miles per hour.
[20:19] Humans are going 20 miles per hour, and we're trying to catch up and understand, uh, what happened and what the agents are doing.
[20:28] And so what what Satya is arguing for here is, and and a great argument, I think, is that we need to understand and have more transparency into how the AI, how the superintelligence is working, uh, the, you know, quote, unquote, chain of thought within these models.
[20:47] So to be able to see, you know, a to b, b to c, what this AI thought about or or what it computed to understand and make certain decisions and then why it made those decisions, and then make sure that we are keeping track of that and making it human readable.
[21:07] Uh, there's something...
[21:09] I I believe it's called neurolease as they call it, and it's essentially how agents or or AI systems speak to each other.
[21:17] It's not really real English, which is a a weird concept.
[21:22] It's, uh, hard to read.
[21:24] I think it's...
[21:25] It is English, but the the words, when you put them together, it doesn't sound like a a a normal kinda human sentence.
[21:32] So, uh, he's he's suggesting we need English reader, like human readable text to understand what these models are doing, uh, and then, uh, you know, really be able to unravel this at a pace that we can keep up with, uh, the AI advancement and keep up with the agents, um, especially as we wanna continue to keep these guardrails in place, continue a safe environment for AI to develop.
[21:58] I think this is a great call out.
[22:00] Uh, it's something that we should take really seriously.
[22:02] And it'll be interesting, I wonder, as we've seen kind of standard protocol come out around AI detection in text or image or video where we have different companies coming out with, uh, different protocol that allows us to detect fraud in those areas, It'll be interesting to see if we come out with safety protocols that are widely adopted where everybody says, hey.
[22:28] You know, without revealing our IP and how our models work in a unique way, We want to make sure that these are safe and that they can be audited and understand it.
[22:41] You know, it...
[22:42] It'll be interesting to see, uh, where that goes, and I imagine that somebody's going to release something.
[22:49] And, hopefully, it's not on the heels of a major safety event or cybersecurity event or something like that.
[22:56] Hopefully, we do this proactively and make these advancements without, uh, kinda having the hand force.
[23:03] So, hopefully, the frontier companies are working on this and supporting each other in this.
[23:08] I think the the AI accord, uh, you know, a few weeks ago at the White House is helpful in moving this, uh, kinda collaboration forward, but we'll see.
[23:18] Uh, because we've seen many letters kind of like this come out from, uh, executives at technology companies in the past several months.
[23:28] You know, Jensen Huang, we've seen stuff from from Anthropic.
[23:32] And so it it...
[23:34] It'll be interesting.
[23:35] Does everybody come together and decide on, hey.
[23:38] This is the protocol.
[23:39] This is what we agreed to.
[23:40] Let's help each other out here.
[23:41] Um, or are we gonna continue to kinda address the issue and then not do anything?
[23:47] Uh, hopefully hopefully, they act on something.
[23:50] Um, for the audience, you know, is this, uh, you think, is this a safety mechanism, or do you think Microsoft is positioning against its rivals?
[24:02] We will see.
[24:06] Apple is hiring the team behind Hux, a startup whose AI host made personalized podcasts.
[24:13] A European commission filing shows the iPhone maker Apple can offer its staff jobs and gets a nonexclusive license to the tech.
[24:22] The founders came from Google's Notebook LM team.
[24:26] The app, uh, actually shut down back in May.
[24:30] Uh, Apple controls a top podcast app.
[24:33] AI made shows could sit right beside human hosts.
[24:38] The filing describes job offers and a license, not a purchase of the company.
[24:43] So this is another acquihire type example disclosed in an EU digital markets act filing.
[24:50] Mac rumors and TechCrunch reported it October ninth and tenth.
[24:54] The filing doesn't say who Apple hired or what it plans to build, just that an acquisition or acqui hire occurred.
[25:03] Again, we keen...
[25:04] Continue to see this more and more.
[25:05] I I I think as smaller companies have success in one realm or another, uh, the actual valuable piece are the people more so than the assets or IP of the company.
[25:18] And so I I believe we'll continue to see more actions like this as, uh, different teams have success from a technological standpoint.
[25:26] When you've got really good talent and you can skyrocket, uh, the the the the actual, you know, limiting asset right now is not the individual IP or the technology itself.
[25:40] It's actually the talent.
[25:41] And so we're seeing more so talent, uh, limitations or or, um, a shortage in the the talent pool, and, uh, interesting to see kinda Apple making an acquisition here.
[25:54] I...
[25:54] I'm curious to see where where this angle takes them from, like, the AI podcast standpoint.
[25:59] Uh, how does this impact podcasts and, you know, human hosts moving forward?
[26:05] We we would be curious, but how does this advance things and and and how will audiences respond to AI hosts?
[26:12] And and then how does Apple integrate this?
[26:14] And it'll be interesting to follow along with that story.
[26:19] I I, again, I, uh, think back to, uh, another talk that we saw from a famous technology CEO just last week.
[26:29] Jeff Bezos was speaking about future in which the cost of goods is so low, uh, and the the the the supply is so high, uh, that we can have a very high standard of living on a very low cost.
[26:47] And his argument here was that we would have a labor shortage because you could have an awesome life without having to work, seeing, you know, single income families, uh, working, you know, less than five days a week.
[27:00] And so trying to get people to work would be the bottleneck in the future, which is an interesting concept.
[27:07] And maybe we see some of that now already with so many acqui hires being announced.
[27:13] The value is on the talent, and there's maybe a lack of talent or the supply pool there is, uh, is short in some regard.
[27:23] So, uh, congrats to the team at Hux.
[27:26] Congrats to Apple, and let's watch this story quickly.
[27:31] Question, is Apple late to AI?
[27:35] We will find out.
[27:38] Robotics and physical AI startups have raised about $48,000,000,000 this year, and they're spending it filming people doing chores and jobs.
[27:48] The Financial Times says labs like physical intelligence and skilled a r r...
[27:53] Are putting cameras in homes, offices, and factories, and running robot gyms to capture the movement and force the open web, uh, never recorded.
[28:05] What is going on here?
[28:06] This is crazy.
[28:07] So data, not hardware, is the bottleneck for robots, and it has to be filmed by hand.
[28:14] PitchBook data shows about $48,000,000,000 raised year to date by robotics and physical AI companies.
[28:21] Uh, NEERA calls real world training data physical AI's scarcest resource and plans five robot gyms by end of twenty twenty six.
[28:33] Footage must be labeled frame by frame, and nobody has said what the people being filmed are paid.
[28:39] Although I've heard interesting reports out of some other data company that, uh, somebody was getting paid $20 an hour to, I think, watch video games being played and and document what was going on.
[28:53] So interesting world here, and we're seeing this more and more.
[28:57] This is talking about, uh, data in the robot space and the hardware space and the fact that this has been a major constraint.
[29:04] A good analogy to help understand this is you think about Tesla self driving cars.
[29:11] You know, Tesla vehicles have been on the roads for, you know, ten...
[29:16] Well, fifteen plus years, really, but the the Model three, I think around ten years with cameras that have been capturing driving data for that whole time.
[29:26] So, uh, you know, billions, I believe, of miles of road data where the...
[29:31] They're filming that data and capturing it to understand what happens in different, uh, scenarios on the road.
[29:39] And Tesla has been able to use that data to train their internal AI models to create full self driving capabilities within their vehicles, uh, meaning that the AI in the vehicle understands what to do in different situations because they've been able to train on this very large dataset.
[30:00] The problem in robotics is we don't have a lot of training data.
[30:05] For instance, you know, we don't have billions of, you know, relative miles of people working and doing things with their hands, you know, folding laundry or cleaning dishes or or vacuuming, you know, cleaning the house or working in a factory.
[30:22] We don't have a lot of video data that you can train these models off of.
[30:28] Another example to help understand this is OpenAI and Anthropic used the entire public Internet to train their models on.
[30:37] Right?
[30:38] So they needed all of the data in the whole Internet that was publicly available to pull that data in.
[30:44] You need massive, massive, massive amounts of data to train these models and and enable them to produce really good outcome.
[30:54] There's a limitation here in robotics.
[30:56] And so we're seeing more companies come out where they're paying people to wear cameras and, uh, provide this video data that they can then train, uh, robots on.
[31:08] And we've seen this for a little while now, but I...
[31:13] In, you know, in the Bay Area, and I think it's expanding across the country, there are companies that will pay you to wear a GoPro, uh, and go, uh, make dinner.
[31:24] Or, uh, I've heard stories where you can, um, you hire a chef.
[31:29] They'll hire a chef to come to your house, make dinner for you.
[31:32] The chef's wearing the camera, uh, and the dinner is free.
[31:36] Right?
[31:36] So they're basically paying because they just want that video data.
[31:40] Um, you know, paying paying, you know, house cleaners to wear these cameras so that they can then train the robots on on that data.
[31:51] It's really interesting stuff.
[31:53] We see it, you know, also not just in video data and robotics and and hardware use cases, but also companies are, uh, their data brokerage has become a pretty hot business, uh, to date where if you are a business owner and you've got proprietary data about how your company works and how that operates, you can sell that data in an anonymized fashion to more of these AI companies because they they need more data to continue to train these models and to get better and better.
[32:30] So, uh, continually seeing this...
[32:32] The...
[32:32] Just thirst and hunger, uh, from AI for more and more data to to make these models better.
[32:40] And we'll continue to see, I think, a demand for this.
[32:44] I don't think this will slow down at all.
[32:46] Uh, in in fact, it it...
[32:47] I think it's only beginning, and, uh, we continue to, uh, see just this desire and this need for for more data.
[32:55] So I guess the curious thing is, um, and and there is a moral implication here as well too.
[33:01] It's like, okay.
[33:03] Let let me pay the, uh, house cleaners to wear the camera to go give me training data that I could then go train a robot that's gonna, you know, replace that person at some point in the future.
[33:15] Uh, interesting moral implications there and and how that that works out, uh, we will see.
[33:22] But I guess that the question is, you know, would you let a robot company film your kitchen, uh, in order to get get paid for that?
[33:32] I don't know that I would.
[33:34] It's interesting.
[33:35] Well, another great show today, folks.
[33:38] I appreciate you guys tuning in.
[33:40] We covered a lot of topics and interesting to see a lot of the develops that are...
[33:46] Developments that are happening really fast.
[33:48] We're having fun over here.
[33:50] It's exciting.
[33:51] A lot of new things in the AI space, and so we wanna continue to touch on this day by day.
[33:57] Um, come here for your AI news, real context, and I'm your human in the loop, Austin Wilson.
[34:03] We'll see you tomorrow morning.
[34:07] TBD, I will make an announcement, but we may, uh, move the show back a little bit tomorrow to try to catch normal people that are awake.
[34:18] Uh, and and so I'm...
[34:20] I might do the show at 08:30 or or 9AM tomorrow instead of our normal 6AM time and see how that goes.
[34:27] But if nothing else, thank you guys for tuning in.
[34:30] I appreciate it, and we'll see you tomorrow.