October 7, 2026
2026-10-07
https://www.youtube.com/watch?v=zEKQ3key8GQ

[0:12] We go live in ten.
[0:26] Every morning, the future moves faster.
[0:31] The models, the money, the power, and the people building what comes next.
[1:10] From Austin, Texas to the world.
[1:17] AI News, Real Context.
[1:20] This is the Austin Wilson Show.
[1:23] I'm Austin Wilson, your human in the loop.
[1:32] Welcome to the Austin Wilson show where we're breaking the AI headlines that matter to you today.
[1:39] I'm Austin Wilson coming to you live from the AI capital of the world, Austin, Texas.
[1:44] Today, OpenAI's hidden model claims to solve nearly 400 math problems.
[1:50] SpaceX is reportedly borrowing $40,000,000,000.
[1:55] Meta, Walmart, Stripe, write rules for AI shoppers.
[1:59] Europe's Mistral unveils a 1,000,000,000,000 parameter open model.
[2:04] Lambda reportedly raising $4,000,000,000 before its IPO, and Google upgrades its nano banana image AI.
[2:13] Let's get into it.
[2:15] First up, this is amazing.
[2:17] OpenAI says an unreleased model produced 372 new math results, including a claimed solution of four dimensional Kikiya conjecture, nearly all from a single prompt.
[2:29] OpenAI posted 722 manuscripts and 372 families to GitHub Tuesday evening and says each one resolves or makes real progress on an open question in math or theoretical computer science.
[2:45] The model was given about 4,000 problems, and the average result used about three hours of ChatGPT Pro thinking.
[2:53] Last month, it's Navier Stokes claim took a swarm of about 10,000 agents.
[2:59] This time it says one agent nearly did all of it.
[3:05] If it holds up, AI is now producing research results faster than mathematicians can check them.
[3:13] Fields medalist, Terrence Tile, has criticized the pace.
[3:17] OpenAI says nearly every result came from one prompt to one agent and says the model is coming soon.
[3:25] Many of the proofs are checked in lean, a proof checking language, which Scientific American says makes them all but certain to be correct.
[3:35] Uh, OpenAI hasn't released the model, the exact prompts, or the compute per problem.
[3:41] MIT's Andrew Sutherland says to treat claims of one shotting problems with a single, uh, agent as unverified until they've actually been tested by humans.
[3:55] Man, uh, this is a phenomenal advancement and and an amazing release from OpenAI.
[4:02] Maybe to counter all of their security woes that they've been having recently, they released 372 claims to to solve some very, very, like, very major math problems.
[4:23] Right?
[4:23] So these...
[4:24] A lot of these problems humans have been working on for decades and even up to a century in some case where we have been trying to make progress on these, have not made...
[4:36] Made little progress, you know, little solves here or there or incremental progress, but no major breakthroughs.
[4:45] Uh, and so for OpenAI to release, uh, claims to solutions for 372 of these is mind blowing.
[4:55] Now, uh, the direct application of what does this matter to the everyday person?
[5:01] Does this change anything immediately, or what's the direct application?
[5:06] There really isn't a lot of direct application here.
[5:10] It's it's not like a math problem that they solved is somehow gonna make our cars more efficient or, you know, make computers that much smarter or anything like that.
[5:21] The the key breakthrough here is kinda twofold.
[5:24] Number one, uh, the fact that the the AI models are able to solve these problems is a huge just, uh, breakthrough from an understanding perspective that they are able to, uh, solve at this volume and at this difficulty is really important moving forward and just opens up the, uh, kinda aperture of the the possibility of what AI can go solve in the future and and the fact that these are novel solutions.
[5:57] So this is not a human, uh, you know, saying, write me a poem for my niece's birthday and and the AI comes up with a cool rhyming thing.
[6:06] No. This is like...
[6:08] They pointed it at 4,000 different problems and said, go figure out which ones you can solve.
[6:12] Right?
[6:13] So that's a that's a big breakthrough from that standpoint.
[6:16] The other notable thing here is the fact that it took on average three hours of a ChatGPT pro, which is a consumer grade model, you know, three hours of pro level thinking.
[6:34] Of course, this is an unreleased model, so this model is gonna be more advanced.
[6:38] But just the perceived compute power that it took to solve each of these, um, is very small, especially, you know, when we look back at the Navier Stokes solutions, you know, just several weeks ago, that took 10,000 agents and eighty eight hours to solve.
[6:58] And now we're seeing agents solve these much faster, which...
[7:03] With much less compute, which in and of itself is a phenomenal breakthrough.
[7:07] And then and then maybe the third piece that goes with this is just seeing that, yeah, Navier Stokes was maybe a month ago or a little over a month ago, and and we've advanced this much in that short of period of time.
[7:21] And so just the rate of advancement is another phenomenal thing.
[7:24] So, you know, big congrats to OpenAI and this release here, and we hope to see more of this.
[7:31] This is really exciting.
[7:34] And and look forward to seeing kind of the solutions that actually do have a a positive impact on the day to day.
[7:42] SpaceX is reportedly lining up $40,000,000,000 of debt led by Apollo.
[7:48] This is in order to buy Nvidia chips for its AI data centers.
[7:52] The Financial Times reported Tuesday that the roughly $40,000,000,000 package would be about $10,000,000,000 of bank loans and $30,000,000,000 of investment grade bonds with Apollo organizing it and bond giant PIMCO among the firms in the talks.
[8:10] It may not close until 2027.
[8:13] Old Elon Musk has said xAI's closest to data center could double its Nvidia chips by December.
[8:21] Uh, that's amazing.
[8:22] AI is now being built on borrowed money or continues to be built on borrowed money, like rail roads and power plants, uh, and a lot of the Internet build out, and we've seen this across technological revolutions.
[8:38] Uh, Morgan Stanley estimates AI infrastructure needs about $1,500,000,000,000 of outside financing by 2028.
[8:47] Reuters reports SpaceX plans to use Nvidia chips exclusively in its future data centers.
[8:54] That's interesting.
[8:55] Per Reuters, SpaceX shares fell 1%, Nvidia rose half a percent in extended trading on the report.
[9:02] Debt gets paid back whether or not the AI pays off, and this deal may not even close until 2027.
[9:09] We love to see it.
[9:11] We love to see putting the chips on the table and maybe borrowing some chips to put on the table.
[9:15] This is amazing from the the Elon companies from SpaceX AI build out.
[9:24] To give some context here, um, Elon Musk and the the SpaceX AI team, uh, went and build out...
[9:34] Built out their, uh, initial data center classes.
[9:37] I believe it's up to a 200,000 GPU cluster, which is massive, massive, massive amount of compute.
[9:47] They they, of course, did not launch the data center at that scale.
[9:52] Um, it it took a little bit of time to to build up to that, but they were able to to start from nothing and have a data data center in a very, very short period of time.
[10:02] I believe it was a hundred and sixty two days or or somewhere around that.
[10:05] Don't quote me on that, but it was a very short period of time to go from breaking ground to having a running data center.
[10:13] And Elon's just running at this full steam ahead, um, and seeing this debt issuance, $40,000,000,000 to go grab more chips and continue to build out.
[10:26] Um, it...
[10:27] It's kind of an interesting thing here where we see at the the top of the stack right now, kinda to to explain this in layman's terms, at the top of the stack, we have your application.
[10:39] So, you know, maybe think about we're starting to see some applications come out if you work at a company that uses Glean as an enterprise search agent or, um, maybe you're using some of these, uh, personal assistant, uh, agents like Muse or Grok Bot.
[10:56] Um, those applications are consumer facing, uh, very much at the top of the stack.
[11:02] Then we move down to kind of the middle layer.
[11:04] That's where the AI models sit.
[11:07] So we've got, uh, OpenAI, and we've got Anthropic with Claude.
[11:11] We've got Groq, xAI.
[11:13] We've got, um, open source models.
[11:15] We've got different models that are being built out, and those are feeding those applications.
[11:20] Well, of course, in order to develop those models, to train those models, to get smarter and smarter models, as well as to serve those models to the the the front end user, um, they need to run on data centers, and they need to be trained.
[11:37] And, uh, and and so it's interesting to see where does the value go in the stack.
[11:42] Because then at the bottom of the layer, we have, uh, the infrastructure.
[11:47] Right?
[11:47] So this is the data centers.
[11:49] This is the bet that Elon's making that Nvidia with their chips are making, and that we continue to see a lot of money move towards the infrastructure and the compute compute layer where the actual, you know, servers and computers are running to serve the the the rest of the stack.
[12:08] And so seeing this bet be made by Elon, it it does two things.
[12:14] Number one, it's like, okay.
[12:16] Elon's not one for, uh, being reserved when he when he takes risks.
[12:21] He takes very big risks.
[12:22] So that that is, you know, something to be considered, but calculated nonetheless.
[12:27] And so we we should see that he he must see something.
[12:31] They...
[12:32] Their team must see something in order for, you know, Apollo and big financiers to back this kind of move.
[12:41] They they must believe in the project and and what they're doing for Colossus two in order to back that.
[12:49] And so there is some positive signal.
[12:51] There is some, you know, accountability here where where we see, hey.
[12:56] Maybe there is something real here.
[12:58] Maybe there's something real.
[13:00] I'm pretty sure there's something real here.
[13:01] $40,000,000,000.
[13:02] That's what I'm talking about.
[13:03] Um, that being said, uh, we've seen this in past revolutions, technological revolutions with the railroad build out, most recently with with the Internet build out.
[13:14] And and specifically in the Internet's case, um, everybody, you know, really believed it got behind, uh, networking and laying fiber cable.
[13:25] We still use fiber Internet today, and and how do we get the whole world connected.
[13:30] And Cisco was a big company behind this in kind of the late nineties, early two thousands.
[13:36] And you...
[13:37] You'll hear this reference sometimes, but they they overbuilt.
[13:40] They were too far ahead of where the demand would be, and Cisco ended up paying for it.
[13:46] They're...
[13:47] You know, they were one of the top stocks of of the day, and they peaked in, you know, early two thousands.
[13:54] And it took them nearly twenty, twenty five years to recover that same stock price.
[13:59] So this is not without risk, of course.
[14:02] That being said, it looks like the demand is there, and, uh, wish best of luck to Elon, uh, and the SpaceX AI team, and looking forward to more compute, more power, more intelligence coming out of these models.
[14:16] Meta, Walmart, Stripe, and Sierra published an open standard for how AI agents shop and act on your behalf at businesses online.
[14:26] It's called the personal agent protocol using OAuth.
[14:31] This is the sign in technology behind, you know, login with Google.
[14:35] You would grant an agent specific read or write access to an account.
[14:40] The business decides what agents can do and can see what they did.
[14:46] So there's a a kind of a trace there.
[14:48] Shopify, Genesis, Instinct, and Rocket also signed on, and a v 0.1 spec is due later this month.
[14:57] Sierra cofounder Brett Taylor, who also chairs OpenAI, told c n...
[15:02] CNBC, it is kind of chaos until such a standard exists, and I think he's he's right there.
[15:11] Right now, a store can't tell whether it's dealing with you or a bot acting for you.
[15:16] Whoever writes this standard shapes how AI shopping works.
[15:21] The backers say it lets a business verify that a bot is acting for a real person, which is great.
[15:26] Sierra published it Tuesday with the partners named, and Taylor explained it to CNBC.
[15:32] It's on paper only.
[15:33] The spec isn't out yet, and Google, Amazon, Apple, and OpenAI aren't named partners.
[15:39] So what does this this mean moving forward, um, and and what are the implications in the context here?
[15:47] It's interesting.
[15:47] This is, uh, awesome to see this come out, to see these folks working together.
[15:52] If you're not familiar, Sierra is a customer service, uh, agent company, uh, that has blown up.
[15:59] They are, um, they've grown very fast.
[16:02] And and, uh, from what I've heard, uh, their their product and the experience and the the agents that they offer, it's like magic.
[16:10] Right?
[16:10] And so they are creating very, very amazing experiences for their customers and transforming businesses.
[16:17] And this is kind of the next iteration possibly of what this would be.
[16:21] So instead of, you know, customer service agents, you know, handling direct customer issues, as we see these personal agents come out, again, Grok bot, OpenAI's dot, Facebook, Meta's Muse, These agents want to go and act on behalf of humans.
[16:40] And one of the biggest use cases that we're seeing is, hey.
[16:43] I need you to go, you know, buy groceries this week, or I need you to go buy a a gift for my mother-in-law.
[16:51] You know, there there are different use cases that are like, I don't have the time or don't wanna go do that.
[16:56] Can I get my personal agent to go do it?
[16:59] The problem is when you have a...
[17:01] An agent kinda clicking around on the Internet, that that...
[17:06] The Internet is not set up for that.
[17:09] Right?
[17:09] The Internet is actually set up to block that.
[17:10] Right?
[17:11] Uh, we don't want bots crawling around.
[17:14] Uh, that's why you see, you know, captchas and, uh, click if you're not, you know, click if you're not a robot, whatever that kind of stuff.
[17:22] And so what what this sets up here is an actual standard and a protocol that says, hey, you know, Walmart, I I wanna go buy something.
[17:34] My agent wanna...
[17:35] Wants to go buy something from you.
[17:36] This is my agent, and I've approved them to go act on my behalf.
[17:41] Walmart goes, okay.
[17:42] We recognize you as an agent.
[17:44] Agent come in, and you can shop on behalf of Austin.
[17:46] Right?
[17:47] So that...
[17:48] That's the idea here.
[17:50] Um, one of the biggest problems, we've seen this, uh, in a couple different stories come out from Amazon and that they are actively blocking agents from purchasing on on the the Amazon web app or on the, yeah, the Amazon store in in general.
[18:08] The big problem there is Amazon generates a great deal of their revenue from ad spend.
[18:16] And so they have advertisers that are...
[18:19] And you you barely notice it, but, know, when you search for, hey.
[18:23] I need to buy some, uh, light bulbs, you're gonna see, uh, a sponsored search pop up.
[18:29] If you have an agent doing that search, those, uh, searches become less and less valuable to Amazon and to their advertisers.
[18:38] And so there's a kind of a complex issue here.
[18:41] It'll be interesting to see how Amazon and, you you know, you've got to imagine it it says Amazon and and Apple aren't onboard.
[18:50] You you know, is Apple going to want to let agents kinda crawl around their ecosystem and make decisions and purchase stuff as well too.
[19:01] So that it'll be interesting to see if all of the technology companies come together with the consumer brands or retail companies and how they work together.
[19:13] And if we do come up with a single agent protocol, I kinda can't imagine that that we come up with one protocol that everybody agrees on, but it'll be interesting to see if that's the the case.
[19:24] That being said, we're we're moving forward, and and and the the direct impact to you is now you you can obfuscate some of the time that you spent, uh, shopping.
[19:39] You know, my wife is shopping for a new sofa, and it's just can't find the right one.
[19:44] Can't do it.
[19:45] Hey.
[19:45] Can I have my agent go find me a few options and come back and let me know?
[19:49] And then can I tell my agent, hey?
[19:50] Go ahead and purchase that for me.
[19:52] Um, the the amount of time that we'll save here is amazing.
[19:56] And then the interesting, you you know, related impact that we see is on the economy.
[20:00] Like, do we see retail and consumer spending increase by any measurable, uh, number because of these agents?
[20:08] Um, we'll continue to watch this closely, but, great to see the the teams working together and seeing some advancement here.
[20:15] France's Mistral unveiled large four nicknamed Le Chonk, a model of about 1,000,000,000,000 parameters that it calls the most powerful open weight AI outside of, uh, China.
[20:29] It has a little over 1,000,000,000,000 parameters with 49,000,000,000 active per word.
[20:36] The same mixture of experts design as Reflections Beam, which we covered yesterday.
[20:41] Mistral trained it from scratch on 4,000 Nvidia Grace Blackwell GPUs over two months in its own European data centers, and it covers more than a 160 languages.
[20:54] Developers can use it through Mistral's API now.
[20:57] The waits are due October 27.
[21:02] Two days, two big non Chinese open models dropping.
[21:06] The open weight race is no longer China's alone.
[21:10] Mistral says it's the most powerful open weight AI system outside of China.
[21:15] Cofounder, Gemini Llama, says m l four is at the frontier of open weight models.
[21:21] Mistral's own benchmarks say that it scores 62% on the software engineer focused DeepSweet.
[21:31] VentureBeat notes that score doesn't establish an outright coding lead.
[21:35] I don't think.
[21:37] But for open sources, that's a great score.
[21:39] Right?
[21:39] You can't download it until October 27, and it ships under Mistral's own license, not a standard open source one, which is an interesting...
[21:49] We'll we'll see what that actually means.
[21:51] So, again, seeing progress here, seeing progress now out of Europe, which is wonderful.
[21:56] We love competition.
[21:57] We love more model companies in the game, and we love more open source model companies in the game.
[22:02] It's interesting.
[22:03] So in in this story, we we we reference Mistral training on 4,000 Nvidia, Grace, Blackwell GPUs.
[22:11] So those are the chips, uh, 4,000 of those, uh, that they trained on it for several months to generate this model.
[22:20] It has a trillion parameters.
[22:22] That means that, basically, it has a trillion little rules in it that each time in a standard LLM, each time that you ask a question, it's gonna run through all 1,000,000,000,000 of those.
[22:36] In this case, just like with the the model that we covered yesterday, uh, the reflection beam model, um, when you ask a question, instead of running it through a trillion parameters, it just cuts a little slice, 49,000,000,000 parameters, and it runs it through those.
[22:53] That's why we need the compute.
[22:55] That's why we need the data centers.
[22:57] That's why Elon's getting a $40,000,000,000 loan.
[23:01] And that's why we continue to see progress in the hardware space because these models and this intelligence and pushing the frontier of this intelligence is so compute heavy.
[23:12] It requires, you know, in this case, 4,000, uh, Grace Blackwell chips running for two months constantly.
[23:20] Right?
[23:21] It's just constantly running.
[23:22] And then when we get users on the other end, they're gonna be using, uh, the the the chips as well to to actually run the models on them and and use them and ask questions and and uncover, you know, 372 math problems.
[23:35] Right?
[23:36] Not to, uh, overlap too many stories there, but this is what's happening.
[23:41] Um, all in all, congrats to the team at Mistral.
[23:46] Congrats to Europe and and to France.
[23:50] And and, like, we wanna see this everywhere.
[23:53] We wanna see everybody coming out with new models and new open source, new open weight models, which means that anybody can use these.
[24:02] They're they're not restricted, you know, and and and especially from a cost standpoint or from a compute standpoint, this opens up the availability for people to use this across the world.
[24:14] And it looks like we're going to get that towards the end of this month.
[24:18] And same same thing with the reflection model, which is really cool.
[24:21] So congrats to that team over there.
[24:23] Excited for them.
[24:24] Excited to see more come out of Europe and to see more advancements in open source.
[24:31] The AI cloud company Lambda is reportedly raising up to $4,000,000,000 at a $14,500,000,000 valuation.
[24:38] Its last private round before they go public.
[24:41] The round of up to $4,000,000,000 values Lambda at $14,500,000,000, uh, before the new money.
[24:49] Blackstone and Codex are leading it, The Wall Street Journal reported Tuesday.
[24:54] And Lambda is aiming for a 2027 IPO.
[24:58] Its order backlog jumped from $15,000,000,000 in June to $50,000,000,000 in September, anchored by a $35,000,000,000 compute deal with Anthropic.
[25:09] Last November, it was valued at $5,900,000,000, and this is just crazy to see over and over again.
[25:17] So this picks and shovels trade, the companies basically...
[25:22] In in in this case, Lambda rents Nvidia chips to AI Labs.
[25:28] So to the Mistrals of the world, to the reflection beams of the world, to the OpenAI, to Anthropic here.
[25:37] Right?
[25:37] So the companies renting Nvidia chips to AI labs are growing as fast as the AI labs themselves.
[25:44] Lambda is one of the, quote, unquote, neo clouds that rent Nvidia GPUs to AI labs, and Nvidia is an investor.
[25:52] So this is good for Nvidia.
[25:54] The $50,000,000,000 backlog and the Anthropic deal, uh, are what's really driving, uh, this growth and this new funding round.
[26:04] Uh, a backlog is promised future spending, and one customer, Anthropic, accounts for most of it, which does create some risk there, but, uh, justifies the additional spend and and growth needed.
[26:19] Again, this kinda comes back to the question, you know, are we getting overbuilt?
[26:24] It doesn't seem like it at this point.
[26:26] This is the big difference between kind of the Internet age and where we are at currently in that...
[26:32] In the Internet, pets.com didn't have any revenue.
[26:36] And little.coms were going public at, you know, tens of million dollar valuations, whatever, but didn't have revenue or had negative revenue, whatever the case may be.
[26:47] In this case, we see Lambda has $50,000,000,000 backlog.
[26:52] It seems like most of this commit, compute commit is coming from Anthropic.
[27:00] But needless to say, Anthropic is about to go public and open up a bunch of free cash where they will continue to spend on compute.
[27:10] You know, we see this constraint.
[27:13] OpenAI, we've covered a couple things around this where OpenAI has done some things in the past, uh, week or so that would signal maybe they don't have, uh, compute capacity.
[27:23] I brought this up to somebody this week, and they said it's it's not a compute capacity problem.
[27:28] It's a money problem.
[27:28] They don't they don't have enough money to pay for the compute.
[27:31] And so the demand is there.
[27:33] It's all about getting the money to flow in the right direction.
[27:37] The revenue is there for these, you know, neo clouds and for the folks at the bottom of the stack.
[27:45] Right?
[27:45] The infrastructure layer, the chip companies we talked about Etched yesterday.
[27:51] And so there's definitely demand here.
[27:54] Uh, whether or not it's oversized, I think we're so early.
[27:59] You know, we talked about, uh, only 2% or something of consumers pay for, uh, AI services.
[28:07] Man, um, we'll be able to see that grow substantially, uh, over the next year.
[28:15] And I think especially in my estimation as we see these breakthroughs from OpenAI, as we continue to, you know, see open source grow, we we see the pace of of development advancement speed up.
[28:29] I believe that by the end of twenty twenty seven, we're gonna have substantial breakthrough such that the everyday...
[28:38] You know, our our everyday lives will be dramatically impacted in a very positive way.
[28:44] And most of us will not look at the world in the in the same way probably by the end of twenty twenty seven.
[28:50] As you look forward to, you know, the next year, 2028, you know, you look at two years out and it's like, man, you can only imagine what the world is gonna look like.
[29:03] It's gonna be a completely different place, and I'm excited for it.
[29:07] I hope you're excited for it.
[29:08] We're excited for Lambda, and wish them the best of luck on their funding round and developing more compute.
[29:18] Google released Nano Banana 2.1.
[29:21] This is an upgrade to its popular AI image generator and editor built on Gemini 3.6 flash.
[29:29] Nano Banana 2.1 is built on, uh, Gemini 3.6 Flash and is rolling out in the Gemini app.
[29:36] Google AI studio and, uh, the Gemini API.
[29:39] Google says it improves on previous versions across the board.
[29:44] Developers pay about 3¢ for a standard one k image and about 7.6¢, uh, for a four k image.
[29:52] We love to see more images.
[29:55] Image editing is one of the AI features regular people actually use, and this lands, uh, right in the Gemini app, which is really nice, especially coming on the heels of Gemini's Argon release last week.
[30:06] Google says it beat, uh, earlier versions across the board.
[30:11] Alpha Signal reports that the new Nano Banana beats the pro model on editing.
[30:18] Google's release notes, uh, and published API prices, uh, are are where this story is coming from.
[30:26] Reports disagree on price.
[30:29] The decoder says about half the cost of nano banana two, while others say the price is unchanged.
[30:35] And Google's claims are the only benchmark so forth so far.
[30:39] So we'll have to see where where the cost lands on this.
[30:41] But, you know, I was just thinking, do we have enough AI slop out there?
[30:49] I don't think so.
[30:50] We need more slop.
[30:53] Sloptober.
[30:54] There's dang.
[30:58] What is the name of the...
[30:59] I...
[30:59] There's there's a company doing Sloptober, which is all AI generated videos, and they're doing a whole contest around it.
[31:08] Marc Andreessen got behind it.
[31:09] We need more slot, man.
[31:11] We need more AI images.
[31:13] Uh, they are getting better and better.
[31:15] It is phenomenal to see, uh, what these models can do and and just the capability.
[31:21] I think it's really interesting.
[31:22] You know, not very long ago, it was like, hey.
[31:26] Generate this image, and it was not really very close to what you were asking for.
[31:32] Maybe it looked really good, but the the accuracy, let's say, was not that great.
[31:36] Um, that is getting much, much better.
[31:38] And then it's going beyond just image and going to, um, to video, which is, uh, crazy to see when you can say, hey.
[31:47] I need a video, uh, you know, that does x, y, and z.
[31:51] And it goes, oh, you mean this?
[31:52] And you're like, yeah.
[31:53] That's exactly what I wanted.
[31:55] And maybe there's, like, a few tweaks.
[31:57] You know, you you do a couple different turns back and forth, but altogether, it's...
[32:02] They're they're turning out really, really good content.
[32:06] So good to see.
[32:07] Nano Banana, congrats to Google.
[32:11] Like, great to see that they are still in the game.
[32:14] I think the image stuff, you know, a year ago with Sora and...
[32:21] I think it's called Sora.
[32:22] Right?
[32:22] The OpenAI one and and Groq, Imagine, and all that kind of stuff was really fun, really cool.
[32:29] That's not where the value lay for, you know, most people.
[32:34] People weren't gonna pay for that, but a really cool tool.
[32:36] But now we'll see.
[32:37] It'll be interesting to see the creative side of things where people are able to take this, which I think is really interesting.
[32:46] Like, just opening up the creativity when you remove the barrier, if you will.
[32:54] Just like with coding, it used to take years to become a really good developer.
[32:58] Now anybody can jump on and say, hey.
[33:00] Build me this app, and it builds an app just the same.
[33:03] It used to take years to become a, uh, really, you know, mature artist, uh, videographer, photographer, musician.
[33:15] Um, now we remove that barrier to to where anybody can develop that.
[33:19] But not only that.
[33:20] This is the the interesting thing is just like in the developing, uh, developer use case and coding, um, uh, examples, with art, there's still a lot of manual, like, not fun stuff that artists have to do to create new art.
[33:39] And so you remove that for the artist, and they become 10 times more prolific.
[33:45] And they know...
[33:46] They have taste.
[33:47] Right?
[33:47] Like, they know what looks good.
[33:49] They know what sounds good.
[33:50] They know the little nuances and details that they need to fix.
[33:53] Now they can fix those and build build these really cool art pieces very quickly, which I think is...
[34:01] That's the exciting thing to me is is not the slop.
[34:04] Of course, I I joke about that.
[34:06] It's not just massive amounts volume of content, but it's where actual artists...
[34:14] Just like in the developer world, we see developers coming up with amazing new ideas that would have never been capable that the average person is never gonna think of.
[34:21] Same thing in the art world.
[34:23] I believe that we're gonna continue to see artists use these tools to develop more and more and better art.
[34:31] You think about Picasso did create an art piece at least once a day.
[34:37] Right?
[34:37] Like, he did one art piece a day, extremely prolific and wonderful gift to the world that that we all get to share.
[34:45] And so we'd love to see that in the the neo artists.
[34:48] That being said, I think traditional, uh, straight up analog human artwork is going to, uh, only increase in value as we see, uh, these, uh, kinda AI supported art pieces come out.
[35:06] So we had a great show today, man.
[35:09] That was our last story.
[35:10] I love this.
[35:11] We're having so much fun.
[35:12] I hope you guys are getting some value out of it.
[35:15] I love talking through AI News, what's important, why it's important to you.
[35:20] Hopefully, you guys are learning something.
[35:22] I'm learning something every day, which is really fun.
[35:25] We're trying to make this show a little bit better one day at a time.
[35:29] If you've got any feedback or anything that you think we could do better, uh, send us a note to taWS dot live.
[35:37] Um, you can also sign up for our newsletter there, The Loop, where we, in written form, give you some more detail, and we'll have more kind of practical application, hands on stuff that everybody can do and and use to to learn this craft and and understand AI and become well versed in it.
[35:54] And that's our whole goal is that you would become well versed in these tools and these capabilities, uh, and that it would impact your life positively and for the better.
[36:05] That being said, you come here for AI news, real context, and I'm your human in the loop.
[36:12] Austin Wilson will see you tomorrow at 6AM central, and we can't wait for it.
[36:19] Let's go.