[
  {
    "i": 0,
    "speaker": "Speaker 1",
    "text": "a confidential virtual machine that runs on qualified computers, right? Running Intel TDX needs like an Intel Xeon CPU. This is a data center level CPU, not something you you stick in like a gaming computer you"
  },
  {
    "i": 1,
    "speaker": "Speaker 1",
    "text": "buy at Best Buy."
  },
  {
    "i": 2,
    "speaker": "Speaker 1",
    "text": "And it also needed a data center level graphics card from Nvidia. So AMD doesn't their their GPUs or their MI I think they make data center GPUs, too, but those are not supported either cuz"
  },
  {
    "i": 3,
    "speaker": "Speaker 1",
    "text": "they don't have the ability to encrypt the GPU memory."
  },
  {
    "i": 4,
    "speaker": "Speaker 1",
    "text": "Um just to kind of not try to explain this with so much jargon. You can think about what we're doing here as if you were a landlord or you're trying to rent an apartment from a landlord. Um but you"
  },
  {
    "i": 5,
    "speaker": "Speaker 1",
    "text": "don't want your landlord to go into your apartment, right? Or see anything about it. And so essentially what confidential computing and TDX allow you to do is prove that the the apartment is up to code. It has"
  },
  {
    "i": 6,
    "speaker": "Speaker 1",
    "text": "all of the standards. It's got a bathroom. It's got the things that you're looking for, but the landlord never actually has to enter the room."
  },
  {
    "i": 7,
    "speaker": "Speaker 1",
    "text": "And so that's really one of the benefits to to what we've been doing on with Targon VM."
  },
  {
    "i": 8,
    "speaker": "Speaker 1",
    "text": "But and it's it's been working, right?"
  },
  {
    "i": 9,
    "speaker": "Speaker 1",
    "text": "Like it works really well, but there's a problem with doing it this way because the miners themselves actually start up the VMs. So that means that in order to keep it secure for multi-tenant, like you know, letting"
  },
  {
    "i": 10,
    "speaker": "Speaker 1",
    "text": "other people use it, we actually have to go down a level and actually a rent out those Docker containers instead of the VMs themselves. Now, I don't know raise your hand if you used Targon before."
  },
  {
    "i": 11,
    "speaker": "Speaker 1",
    "text": "Oh well, we got some we got some Targon enjoyers in here. That's good to see."
  },
  {
    "i": 12,
    "speaker": "Speaker 1",
    "text": "Um you know, if you've if you've used Targon before, you know that you've used containers."
  },
  {
    "i": 13,
    "speaker": "Speaker 1",
    "text": "Um you it's a lot like runpod if you if you haven't used Targon, but you have used runpod. Um and there are problems with doing it this way. Containers are designed to be ephemeral."
  },
  {
    "i": 14,
    "speaker": "Speaker 1",
    "text": "Doing encrypt you know, encrypted backups is very different because it's expected that the the way the Docker is designed is it's expected that you're the one actually running the it's single"
  },
  {
    "i": 15,
    "speaker": "Speaker 1",
    "text": "tenant essentially. Um and so this is why a few months ago and San Francisco hosted by our friend over here, we introduced Targon OS, a confidential operating system that actually locks down the total machine"
  },
  {
    "i": 16,
    "speaker": "Speaker 1",
    "text": "itself, right? So when you start up it was actually so cool. I had it we we I was looking at the demo of it yesterday where basically the only thing that the miner sees is how much emission they're"
  },
  {
    "i": 17,
    "speaker": "Speaker 1",
    "text": "earning and everything else they're not there's no user there's no way for them to enter the operating system at all."
  },
  {
    "i": 18,
    "speaker": "Speaker 1",
    "text": "It's completely locked out and this allows us to actually essentially access the bare metal and create an orchestration system on the bare metal meaning that we can run confidential"
  },
  {
    "i": 19,
    "speaker": "Speaker 1",
    "text": "virtual machines for each tenant themselves. So that allows us to do encrypted diff backups to where we can back up all of your data without seeing anything. Um and and so we're we're"
  },
  {
    "i": 20,
    "speaker": "Speaker 1",
    "text": "we're really excited to be releasing this here the in the next week um in production."
  },
  {
    "i": 21,
    "speaker": "Speaker 1",
    "text": "Um and so some of the things that we've accomplished this year with targon.com is we've deployed over 16,000 confidential workloads growing 300% um and we've also increased our ARPU"
  },
  {
    "i": 22,
    "speaker": "Speaker 1",
    "text": "from $83 to five over to $5,000 which is a 5900% increase in revenue. So yeah, we actually just had our first profitable month last month. Uh you know, >> [applause] >> we bought back. Yeah, it it was it's"
  },
  {
    "i": 23,
    "speaker": "Speaker 1",
    "text": "very cool. Uh we bought back more than we spent on our miners. So we're we're really happy about that."
  },
  {
    "i": 24,
    "speaker": "Speaker 1",
    "text": "Um we also raised uh 10 and 1/2 million to ensure that we didn't have to sell any of our alpha token to fund operations. Um you know, that's that was it's been helpful for us. We've also"
  },
  {
    "i": 25,
    "speaker": "Speaker 1",
    "text": "supported over 20 startups with this technology and we're signing uh multi-year annual or or annual or multi-annual or multi-year uh enterprise contracts."
  },
  {
    "i": 26,
    "speaker": "Speaker 1",
    "text": "Um Now, you know, we've basically solved being able to make sure that your data is secure inside of of Taragon virtual machine. But, [snorts] the thing is is that this doesn't actually kind of solve"
  },
  {
    "i": 27,
    "speaker": "Speaker 1",
    "text": "what the original goal was in '23, uh which was that like we wanted millions of people contributing to this market, not just like five companies that own all the compute. And so, this actually brings us to what we're"
  },
  {
    "i": 28,
    "speaker": "Speaker 1",
    "text": "announcing here today, which is the uh Taragon Taragon Tower Pro, which I'm going to be doing right here."
  },
  {
    "i": 29,
    "speaker": "Speaker 1",
    "text": "Yeah."
  },
  {
    "i": 30,
    "speaker": "Speaker 1",
    "text": ">> [applause] >> This is >> [cheering] >> We've been We've been working on this for for quite some time. Uh this is actually the the just the the case here."
  },
  {
    "i": 31,
    "speaker": "Speaker 1",
    "text": "I We didn't want to deal with export controls on the GPUs and the fact that one of the you know, an H200 Taragon Tower Pro costs the same as my house."
  },
  {
    "i": 32,
    "speaker": "Speaker 1",
    "text": "So, >> [laughter] >> uh so, essentially, what this allows us to do is allow people to rent to to uh purchase and people who are purchasing AI home labs right now, um you know, they're they're limited to"
  },
  {
    "i": 33,
    "speaker": "Speaker 1",
    "text": "what they can do at the wall."
  },
  {
    "i": 34,
    "speaker": "Speaker 1",
    "text": "Um so, you know, you only have four GPUs, six, seven GPUs. This comes with eight GPUs in all of its configurations, um and it comes in a a couple different flavors."
  },
  {
    "i": 35,
    "speaker": "Speaker 1",
    "text": "Um but, one of my favorite parts about this is that when you're not using it as your home lab, you can flip a switch on the back here and it spins up Taragon OS and immediately allows you to sell it"
  },
  {
    "i": 36,
    "speaker": "Speaker 1",
    "text": "back to the Taragon network completely permissionlessly. So, uh we're really excited about that as well. Yeah, it's one of my favorite aspects of this."
  },
  {
    "i": 37,
    "speaker": "Speaker 1",
    "text": ">> [applause] >> It comes in a few different flavors. Um these are the confidential computing supported, so things that you could permissionlessly onboard onto the subnet."
  },
  {
    "i": 38,
    "speaker": "Speaker 1",
    "text": "It comes in an H100 and an H200 variant along with an RTX Pro 6000 Blackwell Edition as well."
  },
  {
    "i": 39,
    "speaker": "Speaker 1",
    "text": "You can go on the website right now dot com slash Tower and and and it's available now so you can just you plug it into a stripe you send a wire transfer and we ship it to your house."
  },
  {
    "i": 40,
    "speaker": "Speaker 1",
    "text": "It's pretty cool so you know from someone else's computer to your own. We need to take back compute you know come and take it."
  },
  {
    "i": 41,
    "speaker": "Speaker 1",
    "text": "Appreciate it everyone thank you guys so much. I really appreciate >> [applause] >> Very exciting."
  },
  {
    "i": 42,
    "speaker": "Speaker 2",
    "text": ">> [applause] >> All right."
  },
  {
    "i": 43,
    "speaker": "Speaker 2",
    "text": "You're going to raffle that at the end of the night."
  },
  {
    "i": 44,
    "speaker": "Speaker 2",
    "text": "No no he said no no they're not doing that."
  },
  {
    "i": 45,
    "speaker": "Speaker 2",
    "text": "We have two panels short panels very exciting panels before our networking hour which will be in the same room."
  },
  {
    "i": 46,
    "speaker": "Speaker 2",
    "text": "We're going to magically transform this room into a beer and wine bash at the Loews so stay tuned for that. The next panel we need the chairs."
  },
  {
    "i": 47,
    "speaker": "Speaker 2",
    "text": "So everybody bring your own chair."
  },
  {
    "i": 48,
    "speaker": "Speaker 2",
    "text": ">> [clears throat] >> You heard from all the subnets what they do very very different things."
  },
  {
    "i": 49,
    "speaker": "Speaker 2",
    "text": "The next panel is about how the subnets turn their infrastructure into market opportunities. So now we have a little bit of a business talk on how how they do that."
  },
  {
    "i": 50,
    "speaker": "Speaker 2",
    "text": "If you have some empty spots next to you you want to raise your hand. We want to give a little space to the folks in the back that have been patiently waiting to see if you there's There's"
  },
  {
    "i": 51,
    "speaker": "Speaker 2",
    "text": "uh spots here in the front."
  },
  {
    "i": 52,
    "speaker": "Speaker 1",
    "text": "And I think we're ready to go."
  },
  {
    "i": 53,
    "speaker": "Speaker 1",
    "text": "We're ready to go."
  },
  {
    "i": 54,
    "speaker": "Speaker 2",
    "text": "Here, I'll give you this."
  },
  {
    "i": 55,
    "speaker": "Speaker 2",
    "text": ">> Oh, thank you."
  },
  {
    "i": 56,
    "speaker": "Speaker 2",
    "text": ">> This panel can introduce themselves."
  },
  {
    "i": 57,
    "speaker": "Speaker 2",
    "text": "Thank you very much. Let's do it."
  },
  {
    "i": 58,
    "speaker": "Speaker 2",
    "text": ">> All right, you ready to go? Here we go."
  },
  {
    "i": 59,
    "speaker": "Speaker 3",
    "text": ">> Hello everybody."
  },
  {
    "i": 60,
    "speaker": "Speaker 3",
    "text": "So, welcome to the Bittensor Subnet panel. My name is Mark Jeffrey. I am a partner with Stillmark Capital and we invest in Bittensor subnets."
  },
  {
    "i": 61,
    "speaker": "Speaker 3",
    "text": "And before we meet our panelists, I'm going to sort of level set with a very brief explanation of how Bittensor works for those who don't know."
  },
  {
    "i": 62,
    "speaker": "Speaker 3",
    "text": "I generally find that there's probably less knowledge about this little universe in which we live than we would like. So, I'm just going to take it from the top and I promise to be a little"
  },
  {
    "i": 63,
    "speaker": "Speaker 3",
    "text": "quick about it."
  },
  {
    "i": 64,
    "speaker": "Speaker 3",
    "text": "Bittensor takes Bitcoin's mining side and makes it programmable. That's the main idea. And Bittensor's main coin is called Tao. And the core idea of Bittensor is let's do Bitcoin again, but"
  },
  {
    "i": 65,
    "speaker": "Speaker 3",
    "text": "this time let's mine for AI. Let's mine for intelligence. Let's do something useful with all this GPU. And for that reason, we call this proof of useful work. And just like Bitcoin, the Tao"
  },
  {
    "i": 66,
    "speaker": "Speaker 3",
    "text": "coin has a 21 million coin cap. And so far about 11 million have been mined to date. So, it's roughly 2013 in Bitcoin terms in in the Bittensor universe. It's roughly equivalent to that era of"
  },
  {
    "i": 67,
    "speaker": "Speaker 3",
    "text": "Bitcoin."
  },
  {
    "i": 68,
    "speaker": "Speaker 3",
    "text": "Now, inside of Bittensor there are 128 subnets. And each subnet is mining a different AI product. And each subnet has its own token. And the subnets to get these tokens you must either mine"
  },
  {
    "i": 69,
    "speaker": "Speaker 3",
    "text": "them or you must stake Tao. So, subnet tokens are thus rehypothecated to Tao."
  },
  {
    "i": 70,
    "speaker": "Speaker 3",
    "text": "And this is an important point. There is a direct economic connection between Tao and subnet tokens. They're basically different forms of the same thing. Now, uh the Tao coin and because of this, the"
  },
  {
    "i": 71,
    "speaker": "Speaker 3",
    "text": "Tao coin basically operates as an S&P 500 of all the subnets. But if you pick the right subnet, you can do exponentially better."
  },
  {
    "i": 72,
    "speaker": "Speaker 3",
    "text": "So, right now the subnet market cap's kind of range from a million dollars up to about a hundred million. And most are sort of floating in the five million to twenty million range. So, very very low"
  },
  {
    "i": 73,
    "speaker": "Speaker 3",
    "text": "valuations for early stage AI companies, right? However, each of these subnets is doing something extraordinary for the most Well, at least the top ones are. A lot of the subnets are doing incredibly"
  },
  {
    "i": 74,
    "speaker": "Speaker 3",
    "text": "extraordinary things, which are on par or better than a lot of the venture-backed AI companies that I'm seeing out there. And we are lucky to have four of the excellent subnet operators up here today. So, why don't"
  },
  {
    "i": 75,
    "speaker": "Speaker 3",
    "text": "you uh each brief briefly introduce yourselves and tell us about your subnet what you're up to."
  },
  {
    "i": 76,
    "speaker": "Speaker 4",
    "text": "Start with Mog."
  },
  {
    "i": 77,
    "speaker": "Speaker 4",
    "text": ">> Uh good afternoon. I'm Mog. Um of Tao Stats, but also um co-founder of Hippias, uh Block Machine, uh GM, and also Videos. I've got a couple of few subnets. Uh mostly infrastructure-based. Uh I skipped"
  },
  {
    "i": 78,
    "speaker": "Speaker 4",
    "text": "machine learning school, and so I'm not really uh pushing the AI narrative as much as the infrastructure that really sits underneath it. That's what I'm doing."
  },
  {
    "i": 79,
    "speaker": "Speaker 5",
    "text": ">> Hi, Mikaela. Um from Metanova Labs, and we run Nova, the Tensor subnet 68, which is a drug discovery subnet. And what it allows us to do is to virtually screen potential molecules, nanobodies, and"
  },
  {
    "i": 80,
    "speaker": "Speaker 5",
    "text": "advance other tools that can help us identify what would be uh novel therapeutic, what would be a successful drug candidate in a decentralized way."
  },
  {
    "i": 81,
    "speaker": "Speaker 1",
    "text": ">> Hello, I'm Robert Myers, uh founder and CEO at Manifold. Um we build a confidential compute cloud uh that uh allows for the permissionless onboard of of new nodes."
  },
  {
    "i": 82,
    "speaker": "Speaker 6",
    "text": ">> I'm Gavin Zants, co-founder of Lead Poet. We host global AI agent competitions around sourcing and qualifying sales leads to provide higher quality leads to sales teams."
  },
  {
    "i": 83,
    "speaker": "Speaker 3",
    "text": ">> Okay, question one."
  },
  {
    "i": 84,
    "speaker": "Speaker 3",
    "text": "Why did you need to do your product as a subnet? So, you know, did this Why did this allow you to create your product better, faster, or cheaper than a traditional startup? Like, why didn't you just raise"
  },
  {
    "i": 85,
    "speaker": "Speaker 3",
    "text": "money like a normal person and hire people, right? Like, why did you do a subnet?"
  },
  {
    "i": 86,
    "speaker": "Speaker 4",
    "text": ">> Supply chain procurement. Uh whatever it is that you from our perspective, whether we're sourcing suppliers for for storage or RPC infrastructure, if we create it as a company, we're limited to the"
  },
  {
    "i": 87,
    "speaker": "Speaker 4",
    "text": "procurement that we onboard uh that's in-house or potentially one or two suppliers. And we're not constantly sourcing to you know, with in any business, if you have a reliable supplier, you tend to stick with them."
  },
  {
    "i": 88,
    "speaker": "Speaker 4",
    "text": "Whereas, the way the Bittensor works is this it's constant competition to drive down the the value of the commodity and the price that we're or the drive up the value and drive down the price. And"
  },
  {
    "i": 89,
    "speaker": "Speaker 4",
    "text": "that's just a constant mechanism flywheel turning, which enables the business to be much much more competitive."
  },
  {
    "i": 90,
    "speaker": "Speaker 3",
    "text": ">> Okay."
  },
  {
    "i": 91,
    "speaker": "Speaker 5",
    "text": "Michaela."
  },
  {
    "i": 92,
    "speaker": "Speaker 5",
    "text": ">> Well, for us it solved a lot of problems simultaneously."
  },
  {
    "i": 93,
    "speaker": "Speaker 5",
    "text": "Uh one of them it it becomes like a self-financing mechanism. So, instead of raising money, you're raising through the protocol."
  },
  {
    "i": 94,
    "speaker": "Speaker 5",
    "text": "But, also as as I was previously saying, in our case, we really believe that, you know, the decentralized adversarial environment is also contributing to us creating better tech and pushing the"
  },
  {
    "i": 95,
    "speaker": "Speaker 5",
    "text": "envelope in in terms of innovation."
  },
  {
    "i": 96,
    "speaker": "Speaker 5",
    "text": "Why Bittensor? I've gotten that question from people that are, you know, crypto savvy, and they're like, \"Okay, why not another Why not another network? And I was like, well, I don't know any other"
  },
  {
    "i": 97,
    "speaker": "Speaker 5",
    "text": "network that has such a talented community that can instantaneously start driving the competition of, you know, that just the the quality of your commodity, of your outputs."
  },
  {
    "i": 98,
    "speaker": "Speaker 5",
    "text": "I don't know how to replicate that."
  },
  {
    "i": 99,
    "speaker": "Speaker 5",
    "text": "Like, sure, you could try, but I just don't know of any other example."
  },
  {
    "i": 100,
    "speaker": "Speaker 5",
    "text": "Um and, you know, for us too, just from like the company perspective, like we didn't have to go through the HR process. We don't need to like look for new talent and we don't need to reward"
  },
  {
    "i": 101,
    "speaker": "Speaker 5",
    "text": "work that doesn't meet our standards. We only reward work that we find valuable."
  },
  {
    "i": 102,
    "speaker": "Speaker 5",
    "text": "So, really what's on us to define is, you know, how do we how do we program?"
  },
  {
    "i": 103,
    "speaker": "Speaker 5",
    "text": "How do we make sure that we can like find consensus and incentivize that quality over time."
  },
  {
    "i": 104,
    "speaker": "Speaker 3",
    "text": ">> So, basically, you have all these people that are building that are that are hunting the pharmaceutical molecules for you from all over the world. So, just like how Bitcoin um what Bitcoin does"
  },
  {
    "i": 105,
    "speaker": "Speaker 3",
    "text": "for stranded energy, Bittensor does for stranded talent. So, you got like all these clever brains all over the earth."
  },
  {
    "i": 106,
    "speaker": "Speaker 3",
    "text": "You don't have to recruit them, they just find you, and then they compete to get your coins, but you don't even pay them, the chain pays them. So, it's not actually coming out of your pocket, at"
  },
  {
    "i": 107,
    "speaker": "Speaker 5",
    "text": "least not initially, right?"
  },
  {
    "i": 108,
    "speaker": "Speaker 5",
    "text": ">> Well, not directly, like it im- it impacts, you know, the value of your token. So, you also want to be judicious about how you're managing that. Um but, there's a different upfront cost. And I"
  },
  {
    "i": 109,
    "speaker": "Speaker 5",
    "text": "think, you know, I I also approach it from like an ideological standpoint in that I truly believe that there are large pools of talent that are on"
  }
]