Cuban Urges Pre-Signature AI Audits to Dismantle Insurer PBM Arbitrage

库班呼吁签署前借助AI审计以破除医保PBM套利 马克·库班指出，自保雇主在签约前使用人工智能审计合同，将通过消除隐性价差，瓦解传统健康险公司和药品福利管理公司的商业模式。 马克·库班详细阐述了Cost Plus Drugs如何通过成本加成的透明定价、与雇主直接签约以及自动化模块化制造舱，来打破不透明的医疗巨头格局。他认为，虽然AI在签约后无法挽救一份根本上存在霸王条款的合同，但雇主和消费者可以在投入资金之前，积极向最前沿的模型输入提示词来审计复杂的付款方协议，并揭露隐藏的价差。库班还概述了他对医疗改革更广泛的构想，包括用带息托管账户取代传统保险费，同时他预见边际药物设计成本将大幅趋近于零，从而赋能医生而非取代他们。 马克·库班主张自保雇主在签署福利协议之前使用前沿大型语言模型进行审计，利用直白的提示词（如“他们到底在哪里坑了我？”）来揭露承运商合同中隐藏的定价差价和免责条款。 (18:51) Cost Plus Drugs正在扩大价值2000万美元的自动化模块化制造舱规模，这些制造舱能够针对仿制注射剂和定制的N对1疗法快速完成换线改装，首批设备已被联邦紧急事务管理局（FEMA）和军方等联邦机构订购。 (5:08) 库班提议进行结构性改革，用由政府担保的银行垫款和联邦灾难性止损安全网支持的个人带息托管账户，来取代商业保险保费。 (49:40) 尽管临床AI正在迅速发展且药物设计成本趋近于零，但人类医生在亲身感官诊断、提供情感关怀以及为患者解读算法见解方面依然不可或缺。 (39:23)

[25:52] Mark Cuban: AI doesn't replace jobs. People who know how to use AI replace the jobs of people who don't know how to use AI. [43:28] Julie Yoo: I think we're going to need more doctors in the future because as care advice becomes more abundant, more people are going to demand healthcare, actual like real life healthcare services. [35:54] Mark Cuban: The cost of designing a drug personalized to you after reviewing your DNA and knowing everything that it knows will start approaching zero as the capabilities of models go up. [28:31] Mark Cuban: AI is not smart. It's, you know, it's book smart maybe, but it's not street smart. It's not, you know, entrepreneur smart. [56:48] Mark Cuban: Don't be shy about using AI in things that feel really uncomfortable and weird because that uncomfortableness is probably the path to something that is groundbreaking and gamechanging.

[0:00] We are not in an efficient healthcare [0:02] market. The economics are broken. [0:04] >> One of the many ills of healthcare is [0:05] that it completely lacks competitive [0:07] dynamics. Just merely by the presence of [0:10] plus drugs, you've actually forced [0:11] change in the incumbents. [0:13] >> United Health 3,000 plus subsidiaries, [0:16] $150 billion in intercomp transfers [0:19] annually. There's often times when they [0:21] will increase premiums, increase [0:23] deductibles, and at the same time force [0:25] you into a relationship with a company [0:28] they already own and you don't know they [0:29] own them. The ultimate mission is to [0:32] disrupt healthcare so that it's [0:34] affordable again. [0:34] >> Is there like some area where you're [0:36] like, why is no one building here? [0:38] >> There's no better time to be an [0:40] entrepreneur than right now. The biggest [0:42] challenge and where AI, I think, is [0:44] going to have the biggest impact is [0:49] Well, Mark, welcome to the podcast. [0:51] Thanks so much for joining. [0:52] >> Thanks for having me on, guys. [0:54] >> Julie, what has struck you about Mark's [0:56] contributions in in the space? What what [0:58] have you always found fascinating about [1:00] about what Mark's been up to with Cos [1:02] Drugs or and just in general, and why [1:04] you invited him on to the show? I mean [1:06] Mark uh first of all just had the [1:08] courage to enter into the morass that is [1:11] healthcare in a way that made such a [1:12] splash and that many many people who you [1:15] know also are in the privileged position [1:17] that Mark is in to just have such a huge [1:19] audience have um been intimidated you [1:21] know to do over over many many decades [1:23] and so um I've loved watching uh the [1:26] market react to what Mark has done in [1:28] healthcare. I think you know the one [1:30] thing I know Mark you and I have talked [1:32] about this in the past but um healthcare [1:33] is like one of the industries like one [1:35] of the many ills of healthcare is that [1:36] it completely lacks competitive dynamics [1:38] and competitive forces free market [1:40] dynamics all that kind of stuff and [1:42] what's been fascinating to watch is just [1:44] merely by the presence of cost plus [1:47] drugs as one one example and we'll talk [1:48] about others that you've done to but um [1:50] just merely by that by by the presence [1:52] of that of that entity and what you what [1:54] you represent and what you're doing [1:55] you've actually forced uh change in the [1:58] incumbents that otherwise like even with [2:00] policy, even with like all sorts of [2:03] other forces like never never changed [2:04] over the course of many many decades. So [2:06] that's that's one of Yeah, cuz I don't [2:09] get to see that stuff cuz I don't talk [2:10] to those guys. [2:12] >> And then I think to your point like even [2:13] absent a contract, I think people are [2:15] recognizing from the demand side of the [2:17] market that they need to show that there [2:19] is some pathway for people to get your [2:21] product, right? And so perhaps even in [2:23] the absence of a direct contract, um you [2:25] know, I do I do hear people talking [2:28] about you know, your solution as part of [2:30] their network. [2:31] >> Really, it truly does not surprise me [2:32] when the big three or even the big six [2:34] use our name as kind of a way to um [2:37] counter the request, you know, you know, [2:40] in the discussion of pricing cuz they [2:42] know that our prices are public, that [2:44] anybody can take our API, and big [2:47] companies that have half a brain are [2:48] pulling that data on a daily basis and [2:51] they're saying, "You have to beat this [2:52] price." And so I I can see that. But, [2:55] you know, I've had at least one company [2:57] and probably more um that have said, [3:01] "Oh, yeah, it's, you know, pick one of [3:03] the big three has told us that you're [3:05] part of their network and we get your [3:07] prices where you're cheaper, and we [3:09] don't have an agreement at all with any [3:11] of the big three." [3:13] >> Mark, let's zoom out and take stock [3:15] about where where you're at with the [3:16] business. You you you launched in 2022 [3:20] uh with a few dozen drugs. In April this [3:22] year, you said you were up to roughly [3:23] 4,000. Uh you're getting into [3:25] manufacturing, hospital contracts, [3:27] potentially AI based care. Why don't you [3:29] take stock a little bit of where the [3:30] business? [3:31] >> I'll give you the rundown. So, we [3:32] launched costblooddrugs.com June 19th, I [3:35] mean January 19th of 2022. 111 [3:37] medications, purely generic, and you [3:40] know, we didn't know what to expect at [3:42] all. I went out and started talking [3:43] about it and it immediately blew up, [3:45] right? And then you know our volumes [3:47] there led the manufacturers to expand [3:50] their portfolio with us and we were able [3:52] to add additional manufacturers. So now [3:55] as you said we have you know 4,000 SKs [3:57] give or take and our business every week [3:59] is up and to the right. So our biggest [4:00] challenge is just keeping up with it. So [4:02] that's segment number one. Then the [4:04] obvious response to mail order is well I [4:07] need to pick things up today. And so we [4:09] created what's called the [4:10] teamcubancard.com. [4:12] I did not name it. And it's basically a [4:15] discount card that allows you to go to [4:17] any of the 17,000armacies that have [4:19] affiliated with us and pick up any of [4:22] our not any of but most of our drugs [4:23] locally because we pay I'm a big [4:26] believer in supporting independent [4:27] pharmacy. So, we pay our affiliate [4:30] umarmacies [4:32] $12 um fill fee, which means that's [4:35] going to make us more expensive for the [4:37] lowest cost drugs because if you have a [4:39] $5 statin plus $12, you're going to get [4:41] it cheaper somewhere else. But at the [4:43] same side, if you really want to support [4:45] your independentarmacies, you can pick [4:47] up the cheap drugs. But on the more [4:49] expensive drugs, the specialty generics, [4:51] the bioimilars were actually much [4:53] cheaper. And for things like bioimilars [4:55] and insulin and things that have cold [4:57] shipping, it's actually cheaper to buy [4:59] it and pick it up locally through our [5:01] team Cuban carb. So that's the second um [5:03] stool in in what we offer. Third is cost [5:06] plus marketplace. Well, before I get to [5:08] cost plus marketplace, let me go back to [5:10] our Dallas manufacturing. So a big part [5:13] of um Alex, my co-founder, Dr. Oxmanski, [5:18] his original cold email to me was he [5:20] wanted to manufacture drugs that were [5:22] generic drugs that are on the FDA short [5:24] supply list because there are pediatric [5:26] cancer drugs. There are things that are [5:28] widely used in hospitals. penov um [5:31] ptocin, epinephrine, this long list of [5:33] drugs and we were able to use AI and to [5:37] use um robotics to shrink what would [5:39] have been a football field or more into [5:42] 5,000 square ft. And um not only that, [5:46] but we were able to turn drugs every [5:49] four hours. So we can make a six-mon [5:51] supply, turn the line, and then go to [5:54] the next drug and continue doing and we [5:56] sell those products to hospitals and [5:57] clinics. an extension of that because [6:00] we've been doing that for you, we [6:02] started working on that five years ago. [6:04] Um, we've been able to take those same [6:06] robotics and AI and shrink it into a [6:10] manufacturing plant, if you will, the [6:12] size of a tractor trailer. And we call [6:14] those pods. And those pods can be used [6:17] for anything and shipped anywhere. The [6:19] government was our first customer. Um, [6:21] we've shipped the first one here [6:22] recently where we just got an order for [6:24] a second and they told us there's more [6:26] behind that because if there's an [6:28] ecological disaster, we can send it to [6:30] FEMA if the if FEMA's around. If there [6:32] is um weather disaster, military [6:35] disaster, god forbid, we can ship them [6:37] there and configure them to make any [6:39] specific type of drug. So we can [6:41] configure a pod for um sterile [6:43] injectables for you know multiple pods [6:46] can be daisy chained together so we can [6:48] scale and or you can have one as being a [6:50] sterile room and another doing the the [6:53] the tablet work for generics and I'm not [6:56] the expert on this. Alex can go into [6:57] this in precise detail, but the thing [7:00] I'm most proud of, and actually Stat [7:02] News just wrote a story on this [7:03] yesterday, the day before, is that we [7:05] can take these pods and configure them [7:08] to do um N of one therapies so that we [7:11] can ship the pod to a research hospital. [7:14] And these these pods cost about $20 [7:15] million. ship it to a research hospital [7:18] and if they're if they specialize in any [7:20] types of therapies that you know for [7:22] those people that are in the N of one [7:24] category they can make it and iterate [7:26] very quickly there and then now we're [7:29] doing these ASO drugs and don't ask me [7:30] what ASO stands for because I can't [7:32] pronounce it um out of our um first [7:35] manufacturing facility in Dallas. So [7:37] that's part two. From that [7:39] manufacturing, we created cost [7:41] plusmarketplace.com. [7:43] Um, sorry, cost plus marketplace. And [7:45] cost plus marketplace takes those [7:47] sterile injectables we make and that we [7:49] also others that we source from other [7:51] providers or other manufacturers and we [7:53] sell them at a cost plus 10 or cost plus [7:55] 12 to clinics, hospitals, providers, [7:57] whoever want to buy them at the cheapest [7:59] price. And we literally can save [8:01] hospitals millions of dollars a year. Um [8:04] but our biggest challenge there is we [8:06] convince the CEOs but we can't always [8:07] convince the purchasing people to do it. [8:09] Um and then finally um because of my own [8:13] experiences as a self-insured employer [8:15] and going around and talking to hospital [8:17] CEOs and CFOs and learning their [8:19] relationship with the big carriers, the [8:22] big vertically integrated carriers and [8:23] all the things they like and dislike [8:25] mostly dislike. I went to them and said, [8:27] "Look, I know you hate um the delays, [8:30] the denies, the preos, the, you know, [8:33] the late pays, the clawbacks, the fact [8:35] that, you know, the your plans are or [8:38] the carrier plans keep on increasing [8:40] deductibles, turning the uh providers [8:42] into subprime lenders because they have [8:44] to collect the deductibles before they [8:46] can offer the care or before they can [8:48] get access to the insurance money. [8:50] Knowing all these things, I'd like to [8:52] work out a direct contract, but I want [8:54] two things. I'm going to tell you what [8:55] I'm going to give you in exchange for [8:56] those two things. One, I'm going to [8:58] trust the doctor. Unless it's something [9:00] crazy, trust the doctor. So, you're not [9:02] going to need a pre-auth. There aren't [9:04] going to be denials. Now, we had [9:05] somebody who had Munchhousen's disease [9:07] that required a denial. But, you know, [9:09] other than that, if it's something [9:10] crazy, I'm going to pay the price that [9:12] we contract for on time. I'm going to [9:15] take responsibility for the patient out [9:16] of pocket. Either I'll pay it or I'll [9:18] I'll take the credit risk for the [9:20] patient itself. So you don't you're not [9:22] a subprime lender, you know, and in [9:24] exchange for all that, I want two [9:27] things. I want a better price because [9:28] your associated administration and [9:30] overhead and doctor time is much less. [9:33] Your yield's going to be higher. And I [9:35] want the ability to take the contract we [9:37] signed, this direct contract between us, [9:38] and publish it on cost [9:40] pluswwellness.com. [9:41] So it can act as an open source for all [9:44] of these contracts so that any employer [9:47] can use them and any provider can post [9:49] their contract and even if you don't [9:51] want to post or use, you can use them as [9:53] templates in your own negotiations. [9:55] >> Well, I love the last piece, Mark, [9:56] because I think that was the genesis of [9:57] our thread back and forth when I [9:59] published a piece a few weeks ago on [10:01] what I'm calling like the great [10:02] replacement cycle of of health plans and [10:04] for I mean for so long so you know [10:06] employer sponsored health benefits is [10:08] the largest class of insurance in our [10:10] country. um if you think about it that [10:11] way. So, you know, roughly about I think [10:13] 155 million Americans are covered by an [10:15] employer sponsored benefit or more. [10:17] >> Yeah, more. But in any case, it's more [10:20] than half the country, 60% of the [10:21] country. And um historically, it's been [10:24] completely, you know, monopolized by the [10:26] buas like these national carriers. And [10:29] um and it's been an unadressable market [10:31] from a startup perspective, right? Like [10:33] it's the the relationships were super [10:35] entrenched. You needed these national [10:37] networks to, you know, be able to play [10:38] in the game. And so the like starting [10:40] cost to sort of enter the arena was was [10:42] too high for startups to get into and [10:44] and the relationships were just too [10:45] entrenched. Now uh the premium uh [10:48] increases are just freaking ridiculous, [10:50] right? Like this year we're seeing even [10:52] amongst our startups um some Buas are [10:55] forcing like 60% year-over-year [10:57] increases in premiums. And it's just [11:00] >> it's even worse than you think, Julie. [11:02] You know, not not to step in, but [11:05] >> because they're vertically integrated. [11:06] So, United Health 3,000 plus [11:09] subsidiaries, [11:10] $150 billion in intercomp transfers [11:14] annually. That would make them number 14 [11:16] in GDP in the United States among [11:18] states. [11:19] >> And so, that tells you how big they are. [11:21] And CVS Sigma are not far behind. And so [11:25] there's often times when they will [11:27] increase premiums, increase deductibles, [11:29] increase um out-of-pocket costs, and at [11:32] the same time redirect you or actually [11:34] force you into a relationship with a [11:37] company they already own and you don't [11:39] know they own them. And you talk about, [11:41] you know, employer based insurance, [11:42] particularly those who self-insure. 60% [11:45] of people who get their health care from [11:47] companies get it from self-insured [11:48] employers. And those employers need [11:51] someone to process all the medical [11:53] claims. put aside pharmacy, the medical [11:55] claims, those things are called [11:56] thirdparty administrators. And if [11:58] they're owned by the big buas, they're [12:00] called ASOS, administrative services [12:02] only. And those administrators, you [12:04] know, we hear all the time about the [12:06] problems with PBMs and how they spread [12:08] price. Well, on the medical side, it's a [12:10] million times worse because [12:13] >> it's also chunk of market. [12:15] >> What's that? [12:15] >> It's also the bigger chunk of the market [12:17] relative to [12:18] >> of course. Yeah, it the medical side is [12:19] much bigger. you know, 10% in growing [12:21] versus the other 90%. And what they do [12:24] is the plans and the hospitals have [12:28] their own contracts. And the hospitals, [12:30] unless they're market dominant, have the [12:33] hate the buas and those contracts, but [12:35] more importantly, whatever that [12:37] relationship is, the self-insured [12:39] employer has no idea what the terms of [12:42] that relationship is. Right? Even though [12:45] it's possible you can go on the hospital [12:47] and look up the tick file, you know, the [12:49] transparency files for that relationship [12:51] and that plan, they're not always [12:53] correct. They're not always right. But [12:55] what ends up happening is the price that [12:58] your the employer pays, the self-insured [13:00] employer pays rarely matches the amount [13:03] that was charged or the amount that was [13:06] paid to the hospital. So there's a [13:08] spread pricing there. And to make [13:09] matters worse, those ASOs that are owned [13:12] by the BUAS make it impossible to [13:15] reconcile the numbers so that you can [13:17] see if you're paying more or less. And [13:20] they do things, you know, there's these [13:21] things called 835 and 837 files, which [13:24] is the data that's sent to the hospital [13:26] and the amount of money that's paid in [13:28] response to those claims. And the [13:30] employers have no access to it. And in [13:32] many cases, they don't even have access [13:33] to their own claims. It is just so [13:36] rigged to your point Julie trying to you [13:39] know remove them and disintermediate [13:42] them is the key the holy grail to what's [13:46] happening in healthcare right now. [13:48] >> Yeah. And we're I mean we're super [13:49] excited to see just an influx of um [13:51] startups that are now able to go after [13:53] this market. A big reason why in [13:55] addition to all these market forces and [13:56] again like the economics just getting to [13:58] a breaking point is the fact that AI can [14:00] play a significant role in enabling a [14:03] full stack like health plan offering to [14:05] be built in a way that reduces all the [14:08] craft that you're describing in a very [14:09] short period of time. [14:10] >> But this all comes down to who takes the [14:11] risk. [14:13] >> While AI I'm a huge AI fan. Live it, [14:15] love it. I think it's important on the [14:17] clinical side. I think it's important on [14:19] the operational side. I think it's most [14:21] important short term on the contractual [14:23] side. But the reality is in a broken [14:26] financial system, even when you [14:27] introduce AI into the economics and the [14:30] operations and the processes, when it's [14:33] already broken, you can't fix it because [14:36] you you brought up the point right at [14:37] the beginning, Julian, you're 100% [14:39] right. We are not in an efficient health [14:42] care market. The economics are broken. [14:45] And you see that [14:48] because if you don't have price [14:50] discovery, that's one thing that [14:52] protects you, that keeps you from having [14:54] an efficient market. But what's even [14:56] worse, and here's where the market, the [14:58] economics really start to fall apart. [15:00] And what makes it tough for AI in the [15:02] short term operationally and and less so [15:05] clinically, [15:06] the contracts are so big and so long, [15:10] the buas through their ASOS get to be [15:12] contractual terrorists. [15:15] And what I mean by contractual [15:16] terrorism, the city of Dallas on their [15:18] PBM contract, I haven't been able to see [15:21] their ASO contract yet, 1335 pages. Mhm. [15:26] >> There's no group of humans with enough [15:28] coffee on the planet to be able to catch [15:30] everything they try to squeeze by. [15:32] >> And so, as a result, no matter what AI, [15:35] once you sign a contract, now hopefully [15:37] you use AI to prevent yourself from [15:39] signing a bad contract. And I think [15:40] that's the lowhanging fruit for [15:42] everybody, but once you sign that [15:44] contract with terms you don't fully [15:45] understand and conditions you don't [15:48] fully you can't even protect yourself [15:49] against, they win. There's no AI in the [15:53] planet that can protect you. And when [15:55] you try to compete with them on plans, [15:58] right, that sounds good and I'm all for [16:00] it, right? What sidecar is doing and so [16:02] many others are doing is really smart, [16:04] but they're having problems because, [16:07] you know, yes, the wraparound networks [16:09] are needed and yes, putting together a [16:10] network is hard. But those buas have so [16:14] many subsidiaries, [16:16] they [ __ ] with you somewhere and you [16:17] don't even realize it. And that's the [16:20] ultimate challenge. And that's why I'm [16:22] so big on the break up big medicine bill [16:24] and going after these companies because [16:26] I want I want this to be an open market [16:29] for every entrepreneur, you know, to use [16:32] AI in every way possible. [16:34] >> And so you you were you you went exactly [16:36] where I was going to go, Mark, which is [16:37] one of the areas where AI is playing a [16:39] role for better for worse is that it is [16:41] is just shining a light on how broken [16:43] the underlying system that humans [16:44] designed is right. So the whole like [16:45] prior bot war situation that's happening [16:48] right now between payers and providers [16:49] is one of you know people blame the AI [16:52] but it's actually no it's like AI is [16:53] just implementing the existing fee for [16:55] service billing system to the full [16:57] extent and showing us how broken that [16:58] system is. So I'm curious to like what [17:00] your take is on what's the end state of [17:02] that like clearly your direct [17:03] contracting uh methodology it sounds [17:05] like is able to reduce a lot of the [17:08] information asymmetry and contract [17:10] clarity and and things of that sort. Is [17:12] that what it's going to take or [17:13] >> so the bot wars in terms of claims and [17:16] up coding and down coding that's [17:18] short-term stuff right that that's just [17:20] like you know okay I need to lose some [17:23] weight I'm not going to eat you know my [17:26] rice I'm not going to eat my frosted [17:28] flakes every day right it's nice on the [17:32] periphery and it's good short-term play [17:35] but it doesn't really change everything [17:36] because AI is a great equalizer [17:40] your AI my AI [17:42] you're open, you know, open AI, [17:43] myanthropic, whatever it is, we're going [17:46] to battle each other out and you might [17:48] win a couple innings, I'll win a couple [17:49] innings, but they're they're going to [17:50] net out at some level. So, I don't see [17:53] dramatic change in the long term. Maybe [17:56] if you're first in, you can pick up some [17:58] things and you see some um lawsuits and [18:00] you see some challenges um when one side [18:03] has it and the other side doesn't, but [18:06] the revenue cycle management on the [18:08] hospital side is is pretty darn good. [18:10] And the Buas have an have a big problem [18:14] in my opinion because they grew through [18:17] acquisitions. We mentioned these 3,000 [18:19] subsidiaries. That's a lot of siloed [18:21] data. And even though you can try to [18:23] find ways to get access to claims and [18:26] specific things that allow you to upcode [18:27] and they're doing that, it's going to be [18:29] hard for them to be become AI native [18:32] companies on the BUA side. And I think [18:34] over the long term they're going to be [18:36] challenged. But all of this starts, and [18:39] I can't say this enough, all of this [18:41] starts before you before those employers [18:44] sign their contracts. If every one of [18:46] your portfolio companies said, "You know [18:49] what? I'm going to run my existing [18:51] benefits contract. If I have a broker, [18:53] if I have a consultant, if I have um a [18:55] TPA, if I have a PBM contract, I'm [18:59] running my existing contract through AI [19:01] and I'm asking and I'm using a prompt [19:03] that asks a single question. Where are [19:05] they [ __ ] me over? what should I do [19:07] about it? And within 30 seconds, you're [19:09] going to see bam, bam, bam, bam, bam, [19:11] bam, bam, bam, bam. And the things I've [19:12] seen running through contracts because [19:14] let me just tell you how much I love AI [19:16] agents. So, I use them all. I use [19:18] Grockbot. I use Eric and I are NBA fans. [19:21] And so, I got the Luca um Donix plant [19:24] because it's $77.69 [19:26] or something. I'm like, this is [19:28] brilliant. Crockpot, $77, the Luca plan. [19:30] I got you. And then I have the Gemini [19:33] plan. And then I have OpenAI's plan, [19:35] etc. etc. And I use the agents and [19:37] instinct too and dot and I use the [19:40] agents and I say you go out and find me [19:43] every state entity organization [19:46] enterprise that um has an open records [19:49] opportunity and I want you to find me [19:52] every single contract RFP whatever it is [19:55] you can find um from those entities and [19:58] I want you to write the letter to [20:00] request it. Put it in an email. Put it [20:02] in my draft. Let me personalize it some. [20:04] tell me when it's there and I'll hit [20:06] send. And it's amazing, right? Stuff [20:09] that's always been at the bottom of the [20:10] list that I always wanted to get to but [20:12] never could. Now it's just like this [20:14] stuff is coming. And so I put it up on [20:16] Google Drive and each of the the [20:18] different um LLMs has access to it. And [20:22] I just tell them all every time there's [20:23] a new um something new loads on the [20:26] drive, you go through and you tell me, [20:28] you give me the good news, you give me [20:30] the bad news, you give me what should be [20:32] done, and then you give me the simple [20:33] version so I can send to the company or, [20:35] you know, school or whatever it is. And [20:37] bam, bam, bam, bam, bam. And that's [20:39] where I get to see all these contracts [20:42] that the big three on the PBM side and [20:44] the BUAS on the ASO side are signing. [20:47] And so I get to see specifically how all [20:50] these enterprises are getting ripped off [20:52] these school districts. Oh my god, [20:55] cities from Denver to Dallas to Fort [20:58] Worth unions. I mean, and if you think [21:02] about it, again, going back to your [21:04] perfect point, Julie, is if that 67%, if [21:08] the employers representing 67%, the 160 [21:13] million plus Americans actually did one [21:16] thing, one thing, ran their contracts [21:20] through AI before they signed them and [21:23] negotiated from that, the entire system [21:27] for the big buas and big PBMs and all [21:30] their subsidiaries would collapse. The [21:32] cost of health care in this country [21:34] would drop like a rock. And so when I go [21:36] and I talk to employers, when I go talk [21:39] to investors and talk about their [21:41] portfolio companies and they're like, [21:43] "Do we need legislation? Do we need this [21:45] that?" I'm like, "No, you need to be [21:48] smart and do your job and bring in [21:51] somebody that I call a healthcare CEO or [21:53] CFO whose only job is to manage the [21:56] contracts and everything associated with [21:58] it." And you will find that your health [22:01] care costs will go down. And then once [22:04] we get you there, then we can talk about [22:06] carving out for direct contracts and [22:07] doing all these other things [22:10] immediately. Like there's a couple um [22:13] public companies that I told to do this [22:15] and came back to me and were like, "Oh [22:17] my god, we had no idea how badly we were [22:19] ripped uh ripped off and I'm begging [22:21] them to mention it on their quarterly [22:22] earnings conference calls and they won't [22:24] do it because they still have to do [22:26] business with those buas even though if [22:27] they left them for whatever reason." [22:29] Yeah. And so if we just do that one [22:31] thing, run the contract soup. Literally, [22:34] if you have more than 1,500 employees, [22:37] the quickest way for or members I should [22:39] say, the quickest way for that startup, [22:41] that tech company, maybe depending on [22:44] how bad you're ripped off, into the [22:45] millions or more, is to just do what I [22:47] just said, run your contracts through [22:48] AI. [22:49] >> Yeah. Mo Mark, maybe maybe zooming out, [22:51] c can you share more about sort of the [22:53] the grand vision for for the business a [22:55] few years out and maybe how you think [22:57] about sequencing to to get there? [23:00] >> So, a couple things. One, I mean, the [23:02] ultimate mission is to disrupt health [23:05] care so that it's affordable again. And [23:08] I'm a fan of universal health care, but [23:09] you can't get to universal health care [23:11] until you have correct data and people [23:14] can make decisions based off of real [23:16] data. So, that's bigger system. whatever [23:20] direction it ends up going, I just want [23:21] to disrupt healthcare so that when [23:23] people think about the economics, it [23:25] becomes okay, I know exactly what's [23:27] going on, how we get there, more [23:30] education than anything else. Like I [23:31] mentioned before in terms of the [23:33] business of cost plus drugs. Look, [23:35] somebody could come along for the retail [23:37] site and say, "We're going to do cost [23:38] plus 1%." Okay, trust is still our [23:42] primary product. That's what I try. We [23:44] don't spend any money on marketing or [23:45] advertising. And when we first got [23:47] started, all the people who came from [23:48] farming said, "Oh, you're going to get [23:49] killed." I'm like, "No, you know, people [23:51] are going to tell others how much money [23:52] they saved." And then they're going to [23:54] trust us. And that's the one thing [23:55] missing in all of healthcare economics, [23:57] trust. And so once we gain their trust [24:00] by doing exactly what we said by being [24:02] fully transparent, only having a set [24:04] markup of 15% will garner people trust. [24:07] So that'll protect us elite to some [24:08] level. But the Grand Dio's vision, those [24:10] pods I told you about that sell for $20 [24:13] million, [24:15] there's good margin in those. And we [24:17] think we can sell within two years, a [24:20] couple hundred a year, within five [24:23] years, a couple thousand a year. So you [24:25] give me, you know, a thousand um pods to [24:28] sell at $20 million each, we're going to [24:31] be doing really, really, really good. [24:34] And it'll completely change healthcare [24:36] because we can daisy chain those things. [24:38] Like literally, we could manufacture [24:41] tableted generic drugs in Dallas, Texas [24:44] using these and make them for less than [24:47] we pay India, the manufacturers we [24:50] partner with in India. Now, we don't go [24:51] for the cheapest price. We go for the [24:53] best manufacturer. So, we pay a little [24:55] bit more. But still, we can make it for [24:57] less. The only thing holding us back is [25:00] that for each generic we have to fill [25:02] out we have to um apply for something [25:04] called another new drug application. And [25:07] each one of those things cost $365,000 [25:10] plus other fees. So you want to do a [25:12] thousand of them that's $365 million [25:14] plus all the cost of the pods. We're [25:16] trying to say look if we truly want to [25:19] um bring manufacturing and onshore for [25:21] medications we're here for it and we can [25:23] do it. you're already working with us in [25:25] other areas and so we think we'll get [25:28] there as a big part of what we'll do and [25:30] we've already planned so that on cost [25:32] plusddrugs.com we'll put a little bit of [25:34] little American flag when something is [25:37] manufactured by us here in the US but [25:39] that's probably two years away just [25:41] because of all the work we have to do [25:42] with the FDA but that's the big picture [25:45] if we're if we're selling $20 billion [25:47] worth of pods a year [25:50] watch out Elon [25:52] >> and and so is your view that as If you [25:53] look out the next few years, no matter [25:55] how good the technology gets, it's not [25:58] going to sort of disrupt uh sort of [26:00] healthcare on its own and that there [26:01] needs to be other more structural [26:03] changes that need to happen. [26:04] >> Yeah. Yeah. There's regulatory capture [26:06] and there's scale capture. Right now on [26:08] the economic side, it's all scale [26:10] capture. And when you listen to the [26:12] cordians conference calls or read the [26:14] transcripts for all those big vertically [26:16] integrated companies, they talk about [26:18] technology, but the one thing they never [26:20] say is this is going to translate into [26:22] lower costs for employers and patients. [26:26] They always say this is going to improve [26:28] our earnings. You know, we're going to [26:30] return more to shareholders. We're going [26:32] to buy back stock. [26:35] But until I'm able to educate all these [26:37] employers, right, they're going to keep [26:39] on making it nearly impossible. Which is [26:42] why, you know, when Julie talked about [26:43] earlier about CVS, why I mentioned we [26:46] don't deal with any of them directly [26:49] because we have to work outside of them [26:51] because otherwise you fall into their [26:52] trap and they gave me with their [26:54] contracts. [26:56] Mark, what do you um in all this, you [26:58] know, you're talking a lot about uh [27:00] structural things, but there is a [27:01] reality of what you described earlier [27:03] like AI will change the nature of [27:04] people's jobs in this industry quite a [27:06] bit. What do you think is I I've heard [27:07] you talk a lot about, you know, how do [27:09] we sort of futureproof our, you know, [27:11] young pools of talent and you know, what [27:13] is um being like sort of career ready [27:15] look like in this day and age? What do [27:17] you think that looks like for [27:18] healthcare? Like what does the the [27:20] workforce of healthcare look like in the [27:21] next 5 years? [27:22] >> I mean, it looks the same as any other [27:24] industry. you have to still hire [27:26] entry-level people [27:28] >> because no matter what the the model is [27:31] the minute it's released that's it right [27:34] it's it's behind it's behind and so [27:36] there's latency associated with every [27:38] model and so you need people who can who [27:41] are aware of everything that's going on [27:42] that create context and deal with the [27:44] models be the human in the loop [27:46] understand the consequences etc and you [27:48] need those people to grow with the [27:49] organization over time so that all as [27:52] all these things changed they've seen [27:54] the history of the implementation and [27:56] they can bring value to the [27:57] organization. So that's part one. You [27:59] can't just get rid of all entry- level [28:01] positions. Part two, you have to enable [28:04] those people and everybody else white [28:06] collar oriented how to use AI because AI [28:09] doesn't replace jobs. People who know [28:12] how to use AI replace the jobs of people [28:14] who don't know how to use AI. So that's [28:16] part two. And then part three is the [28:18] bigger picture stuff. The first step to [28:21] that is trying to educate the seauite on [28:23] how AI even works. Teaching them how [28:26] agents work. Teaching them how to use a [28:27] model. Teaching them, you know, the [28:30] difference between open- source and [28:31] closed models and the things you [28:33] different things you have to do. Helping [28:35] them understand the latency of a model [28:37] that this brilliant conversation the [28:38] three of us is having are having is in [28:41] no model anywhere and won't be until [28:43] it's published. So, if we say something [28:45] that we can all go out and use, the [28:47] models are all behind and AI is all [28:49] behind. So that's a given and of course [28:51] you know LLMs are text and some pictures [28:54] and don't deal with you know all the you [28:56] know real world things that are going on [28:58] in video. So you have to be able to [29:00] understand as a CEO and in the seauite [29:02] how all those things impact your [29:04] company. Once you get a good grasp on [29:06] that then you just start fooling around [29:09] right because there you can just find [29:11] new ways for things to happen that [29:13] benefit your customers or create new [29:15] opportunities. And that's the brilliance [29:17] of AI, right? AI is not smart. It's, you [29:21] know, it's book smart maybe, but it's [29:23] not street smart. It's not, you know, [29:25] entrepreneur smart. You've got to have [29:27] somebody there who guides it and helps [29:29] it. But until you have a seauite that [29:31] understands how it works, it's going to [29:34] be very difficult for that organization. [29:36] I mean, I literally sat with the head of [29:38] a a consumer food company and he was [29:41] talking to me about acquisitions and [29:43] this was months ago and he was like, [29:46] "Yeah, because of GLP1's yada yada [29:48] yada." And I'm like, "Okay, well, let's [29:50] do this. Let's take your number one [29:51] product, your cookie, right? And I'm [29:54] going to ask Chat GPT to create a [29:57] competitor that is more GLP-1 friendly [30:00] and will be priced lower. And I want a [30:02] full bill of materials. I want sources, [30:04] everything manufacturing. what I have to [30:06] do and in about 3 minutes we got back to [30:08] him and he was like oh [ __ ] he had no [30:11] idea [30:12] >> now if you have no idea that's a [30:14] question you can ask an LLM [30:16] you don't have no idea on how to enable [30:19] your company to use AI and that to me is [30:22] the greatest greater struggle for most [30:24] organizations now for entrepreneurs [30:26] asking that question and trying to [30:28] implement no matter what it is there's [30:31] no better time to be an entrepreneur [30:33] than right now you know There will be [30:35] the trillionaire in the basement. There [30:37] will be, you know, somebody who puts out [30:39] their own um agent swarm but does it on [30:42] a um entrepreneurial basis. You know, [30:46] now there's still challenges [30:47] particularly for agents. So agents don't [30:50] see social media posts. They can't [30:53] review them, right? And it's hard to get [30:55] transcripts for any material amount of [30:57] them. They can't, you know, watch ads or [31:00] see ads. And so if we're moving to an [31:03] agent-driven economy for individuals, [31:06] there's going to have to be a lot of [31:08] changes that occur in how we communicate [31:11] commerce to those agents. Like [31:12] literally, I was fooling around with one [31:14] of the guys in my company and we created [31:16] this thing and you can go check it out. [31:17] It's not truly active. It's called [31:19] agentfeeds.dev. [31:21] And the whole idea is you take the [31:22] transcript of your post and you just put [31:25] it into um agentfeeds.dev dev and then [31:27] you copy and paste it for anybody can [31:29] copy and paste, you know, anybody you [31:31] follow into their agents. Hey, follow [31:33] all the feeds for this and it takes all [31:35] the ads and gives you all the [31:36] information you need so your agent is [31:38] more up to date on those categories or [31:40] whatever. You're going to see, you know, [31:42] you're going to see an explosion of [31:44] those types of things, but we still have [31:46] to get through trust and you all the [31:48] things we all know about AI and the [31:49] issues we all know about AI. But in [31:51] terms of opportunities going back to [31:53] health care, the biggest challenge and [31:56] where AI I think is going to be have the [32:00] biggest impact other than clinical is by [32:04] patients by consumers, right? And being [32:07] able to look at the things that you are [32:09] given on the economic side that you know [32:12] uh the benefits bill, right? The EOB, [32:14] right? The explanation of benefits, [32:16] right? And here's where I think um [32:18] consumers will get the greatest benefit. [32:21] So right now there's terror in I don't [32:25] know half the country, right? 40% of [32:27] people in the country can't afford a [32:29] $400 emergency bill. And so what I [32:32] always tell people is before anything [32:34] happens, go into your favorite AI model [32:37] and say get me a list of all the [32:39] nonprofit hospitals. go to go to them [32:42] and ask them what their financial [32:44] support services are because all um all [32:48] nonprofit hospitals have to have [32:50] financial assistance programs and some [32:52] of them are really good and for some [32:54] people you're better off and I don't [32:56] want to make the decisions for people [32:58] but you can ask the questions of the AI [33:00] to say whatever my um insurance plan [33:03] premium was going to be I'm going to put [33:05] it in a money market fund and let it [33:07] collect interest because I've already [33:09] discussed my situ situation with the [33:11] local nonprofit hospital and I know [33:13] everything there is to know about their [33:15] financial assistance plan. Now those [33:17] financial assistance plans are created [33:20] and the pricing is a reference price to [33:22] the federal pover poverty levels in most [33:24] cases but it can go up to 600% of [33:27] federal pro poverty level which can be [33:30] over $100,000 for some families in some [33:32] states. So patients slashconsumers [33:36] slashfamilies [33:37] understanding what their options are and [33:40] how to procure those options from [33:42] particularly nonprofit hospitals is [33:44] going to be enormous. If I was going to [33:46] start a company um in this space, I [33:50] would go to every nonprofit hospital and [33:53] set up a plan that post agents that I [33:57] enable people to use. Hey, you know, if [33:59] you're concerned about how you're going [34:00] to be able to pay for your bills or your [34:01] family bills and you, you know, you [34:03] don't have much money in the bank, [34:04] whatever you're making under X number of [34:06] dollars, copy and paste this into your [34:09] Muse or into your DOT or into, you know, [34:12] your um Claude Co-work, whatever it is, [34:15] and it'll know to come here to us and [34:18] ask questions. And so when I taught the [34:20] heads of hospitals, I'm like, you don't [34:22] understand. You guys suck at marketing. [34:25] You are dependent on the buas and their [34:28] plans and being part of their network to [34:30] feed you patients. That's ridiculous, [34:33] right? You if you make it easy for [34:36] people to use your financial assistance [34:38] programs, maybe go to a local bank, [34:41] maybe show people how to learn use um [34:43] stop-loss for for smaller medium-sized [34:45] companies. So many options for them to [34:48] use it to make things better for [34:50] patients because that's really where the [34:52] big disconnect is. You can offer better [34:54] plans. You can use technology to reduce [34:57] the price and become more efficient in [34:59] how you create a network or negotiating [35:01] direct contracts. But until the end user [35:04] truly [35:06] embraces it and knows how to use it, [35:08] there's going to be challenges. But all [35:10] of us at some level, just like we [35:12] learned to use spreadsheets back in the [35:14] day when we thought it was too [35:15] complicated. I'm old enough to remember [35:16] when people thought spreadsheets were [35:18] hard, you know, and you had to explain [35:20] to them and they had to pay $500 for [35:22] Lotus 123. But the same thing is [35:25] happening now as we go through this [35:27] transition and people, yeah, there's a [35:29] lot of people who are anti-AII, but [35:31] there are a lot of people who are [35:32] anti-cal, anti- everything. That's just [35:35] a matter of time. Um, but once you get [35:37] the way you get them to overcome AI is [35:41] their fear of AI is to show them how [35:43] these are the things that can really [35:45] benefit you and protect you and reduce [35:48] the inherent fear that you have in the [35:50] event that something goes really really [35:53] wrong. [35:54] >> Yeah. There's this um famous Aaron [35:56] Schwarz quote from many years ago. You [35:58] know, information wants to be free. Um [36:00] Yeah. And uh and I know in healthcare in [36:03] a previous conversation you've been a [36:04] bit skeptical that that will happen [36:06] because as you mentioned there's a lot [36:07] of siloed data and there's a lot of you [36:08] know strategic reasons why why that data [36:10] is is is you know that that is their [36:12] their advantage to some degree and and [36:13] so I say that to ask you what do you [36:15] think the ultimate impact of the [36:17] frontier labs you know open AI and [36:18] thropic have have very big ambitions in [36:20] the space either them or others will be [36:22] in in re reshaping uh healthcare or sort [36:25] of you know [36:25] >> I'll tell you exactly I'll tell you [36:27] exactly um and I'm selling my own book [36:29] here a little [36:31] The cost of designing a drug [36:33] personalized to you after reviewing your [36:36] DNA and knowing everything that it knows [36:38] will start approaching zero as the [36:41] capabilities of models go up. Right? We [36:44] talk about GPUs and how they'll get [36:46] faster, the networks behind them, the [36:48] power required, but it's the algorithms [36:50] that are going to change everything and [36:51] make them more efficient. Right? And we [36:53] don't even know what the speed of [36:54] improvement are going to be on [36:55] algorithms. at some point in time, [36:57] whether it's 5 years, 10 years, 20 [36:59] years, they're going to take something [37:01] from Eric, Julie, and Mark, and they're [37:03] going to say, starting with over the [37:04] counter, because if you think about it, [37:06] right, you walk into a drugstore, you [37:08] get this, you know, um, excuse me, [37:12] whatever, Advil, Tylenol, right, [37:14] ibuprofen, and we all think it's going [37:16] to impact us the exact same way. And we [37:20] have to experiment to see how it impacts [37:22] our body. the body makes no sense [37:24] whatsoever. [37:26] And so as the processing power and the [37:28] analytics power of these models from the [37:30] frontier models gets better, they will [37:33] be able to really truly personalize over [37:36] the pro over-the-c counter drugs without [37:38] a lot of the mishagos you have to go [37:39] through to try to figure these things [37:41] out, right? And now all of a sudden if [37:43] here's my personal um ibuprofen, here's [37:46] my personal dose or mixture of whatever [37:48] these over-the-c counter drugs, bam, you [37:50] start to change it. Then you get into, [37:53] you know, making other drugs that are [37:55] personalized for you because again, the [37:57] fact that we all take Eloquis and decide [37:59] between Eloquis or Zerelto or Warfare [38:02] makes no sense whatsoever in the big [38:03] scheme of things. And so being able to [38:06] take those things and personalize that [38:08] medicine. And what I have told two of [38:10] the Frontier monks, I said, "If you [38:12] really want to be smart, you take your [38:15] $20 or your $200 um subscription and you [38:19] include that service once it's available [38:23] because your marginal cost for doing [38:25] this at that point in time, look, the [38:28] inference and everything that has to [38:29] happen just to respond to a request [38:31] today expensive like this, right? Coming [38:34] down very quickly. But the same thing is [38:36] going to happen for this type of [38:38] personalized medicine. Initially it's [38:40] going to be enormous. But as we [38:42] understand the human body more and [38:44] understand all the variables in our body [38:46] more and our brains more etc and [38:48] understand the the medications and [38:50] figure out how to do the end of one for [38:52] everybody. It will be part of a normal [38:56] subscription because there's no reason [38:57] it should be. How expensive is 23 and [38:59] me? You pay $99 or less now to get your [39:02] DNA looked at. And so maybe I pay $99 [39:06] upfront. I give them my DNA if I'm [39:08] willing to do that. You know, obviously [39:10] for privy privacy concerns, they store [39:12] it. They continue to bam bam bam bam bam [39:14] bam. You know, maybe you want to know [39:16] what could be wrong. Maybe you don't [39:17] want to know what could be wrong based [39:19] off of your DNA. But if something does [39:21] go wrong, [39:23] >> they have access to all these drugs, all [39:26] these opportunities, and the ability to [39:28] create them in real time. and selling my [39:30] book. We want our pods to be right there [39:32] to be able to manufacture them because [39:34] we're the only ones capable of doing it. [39:36] And I don't see that changing for a long [39:37] time. [39:38] >> I think one of the things also, Mark, [39:39] that people um probably don't appreciate [39:41] as much as they should in that context [39:43] is that as that happens is what you just [39:46] described happens and as drugs become [39:47] just better products for like every [39:49] individual in the country or in the [39:50] world. Um that is actually one of the [39:52] most disruptive forces for care [39:54] delivery, right? like how much of [39:56] healthare is actually the job of the [39:58] doctor helping patients figure out why a [40:01] drug doesn't work for them and [40:02] calibrating and doing met titration and [40:04] changing behavior to make your your life [40:06] work for the drug as opposed to the [40:08] other way around. And so um that's [40:10] something that I think you know [40:11] >> but remember because of that latency I [40:12] mentioned earlier you're still going to [40:13] need doctors if Mark walks into the [40:16] hospital and I just tried this new [40:18] personalized um medicine and I just [40:20] start puking my guts out. [40:21] >> Yeah, you still need tele. [40:23] >> You can have it on video, right? Then [40:24] you can have a robot there say, "How do [40:26] you feel, Mark?" Right? And the [40:27] Terminator might might walk in and say, [40:29] "Your time is up." I don't know. But but [40:32] I'd rather have a human there with all [40:34] five senses being able to take a look [40:36] and ask questions and see the crying mom [40:39] or the crying son or child or whatever [40:41] and be able to include that in the [40:43] context. And so I don't think ever that [40:46] will replace doctors and we shouldn't [40:47] even try because and and you know, you [40:50] guys know how adoption of technology is [40:52] particularly for this. You still have to [40:54] know the questions to ask. [40:57] >> You still have to know what the [40:58] variables are that matter versus the [41:00] ones that don't. You still have to [41:02] realize that even as processing costs [41:05] drop towards zero hopefully at some [41:06] point in time that you know there's [41:09] still a cost associated with it even if [41:11] that cost is only time. And so you know [41:14] you need somebody there who's been [41:16] educated in all those pieces. I think to [41:18] your point, Julie, where things might [41:20] change dramatically when it comes to the [41:22] the role of the doctor is how they'll be [41:24] trained in medical school. They won't be [41:27] taught the exact same things. They'll [41:29] have to learn how to use open evidence [41:30] or whatever else is available out there [41:32] to and how to extract information even [41:36] if it's just that DNA information and [41:38] understanding what's in there and what [41:39] the model thinks about it and being able [41:41] to translate that to the person because [41:44] you know [41:46] effectively in order for this to work [41:48] perfectly everybody has to have their [41:50] own model and that's going to be [41:53] difficult for everybody even though [41:54] there's going to be millions or not if [41:56] not hundreds of millions or billions of [41:58] models [41:59] created and supported over the next [42:00] years. And plus on top of that, IP [42:03] becomes a totally different problem that [42:07] gets introduced into this equation and [42:09] process because you know up until [42:12] recently it's been publish or perish. [42:14] And if you're a doctor and you create [42:15] something, you want everybody to happen, [42:17] maybe you can even license it and make [42:18] some money off of it. But now, you know, [42:20] if you try to patent something, every [42:22] 12-year-old is saying, "Hey, look at [42:24] this patent, change it just a little bit [42:25] and make my own patent." Right? or if [42:28] it's, you know, a new procedure in [42:30] healthcare, you know, you might say, [42:31] "No, you have to come to my model and [42:33] you have to pay me $9,999 [42:36] every access, right? Or to use [42:38] inference, you know, you're going to use [42:40] this software that I require because [42:43] nobody's going to give away that IP for [42:45] free." And that creates a lot more [42:47] problems. I think part of the [42:49] underpinning that a lot of AI [42:50] entrepreneurs have is that going back to [42:52] Erica, information wants to be free. the [42:55] information wh may want to be free but [42:57] the person who owns the IP feels [42:59] differently and so that's going to [43:01] change how we interact with models and [43:04] you know we can argue about you know [43:07] frontends that choose which model is [43:09] least expensive to answer the question [43:12] but we haven't had the discussion on [43:14] where we find the IP that's needed to [43:17] either train that model post- trainin [43:18] that model you know rag that model [43:21] whatever it may be if that person isn't [43:23] making it available but It's absolutely [43:25] required. Does that make sense to you [43:27] guys or do you guys think I'm an idiot? [43:28] >> You know, I think it's really [43:29] interesting because it also brings up [43:30] the point of like what is the role of [43:31] the doctor in the future? It so number [43:33] one, I agree that actually I think we're [43:34] going to need more doctors in the future [43:36] because as care uh advice becomes more [43:38] abundant, more people are going to [43:40] demand healthcare, actual like real life [43:42] healthcare services. And so we're going [43:43] to find we actually need more doctors. [43:45] And um there's something to be said [43:46] about what you what you're describing as [43:48] like the role of the doctor today is [43:50] very rote, right? It's kind of like the [43:52] these are the tasks you got to do these [43:53] quality measures. It's like you you like [43:55] barely need to use your brain honestly [43:57] for a lot of the things that doctors are [43:58] asked to do in the traditional [43:59] healthcare setting. And what you're [44:01] describing is like what if like the [44:02] paradigm moves towards physicians are [44:04] actually like more physician scientists, [44:06] right? Like it's actually their job is [44:07] to take like the bleeding edge of [44:09] information that we may not know at [44:11] population scale yet, but on an [44:13] individual basis they're at the tip of [44:14] the spear of collecting novel data, [44:16] generating novel insights about how [44:17] people are reacting. [44:18] >> Novel data by definition stimulating job [44:21] description to me. [44:22] >> Yes. [44:23] >> Yes. Because by definition, you're [44:25] exactly right. Right. Novel data equals [44:28] person. [44:29] >> Mhm. [44:30] >> Exactly. And with like data, [44:33] >> we're going to have longitudinal [44:34] monitoring. You know, like the blood [44:36] pressure is the one that like we always [44:37] talk about is where, you know, it's kind [44:39] of the same principle as what you're [44:40] saying. The fact that we take our blood [44:42] pressure like once a year is kind of [44:43] ridiculous because it changes. [44:45] >> Not me. [44:45] >> I got this [ __ ] right here. [44:47] >> You're Yeah, you're cutting edge. And [44:48] you know, but soon we're going to have [44:49] wearables that are just measuring it in [44:51] real time like every single day. And [44:52] like you'll have a much more precise [44:53] understanding of how you're responding [44:55] to different therapeutic interventions [44:56] and even like diet and exercise and all [44:58] that kind of stuff. And you know that's [45:00] going to you bring up a good point Julie [45:02] in terms of novel um information and [45:06] that's everybody in terms of finding [45:09] jobs, right? You're going to need AI [45:12] translators for people who may not keep [45:14] up with these things. you know, regular [45:16] people that are in wherever in this [45:18] country or in the world who just, you [45:21] know, want to live their abundance life, [45:23] whatever that may be. And you're going [45:25] to need people who are going to be the [45:28] translators for what's novel about this [45:30] circumstance or this individual and how [45:32] do you, you know, make it valid to AI [45:35] and the output of the AI valid to the [45:37] novel to the individual. That's going to [45:39] be a job and it's going to be very [45:42] difficult to extract everything um from [45:45] an individual that the AI is going to [45:47] need and you're gonna have need to have [45:48] a person I think there to do it. Maybe [45:51] somebody envisions a humanoid robot that [45:54] is going to ask you those questions and [45:56] somebody's going to be comfortable. I [45:58] don't I don't think I don't think that's [45:59] going to happen. [46:00] >> Yeah. [46:01] >> By the way, Mark, I don't think I've [46:02] ever heard you talk about another set of [46:04] entrenched incumbents, which is the EHR [46:06] companies. [46:07] >> Uhhuh. What what's your what's your [46:08] theory of the case there on how that [46:11] that market changes or evolves or [46:13] doesn't um in the age of AI and what [46:15] that looks like in the next few years? [46:17] >> Honestly, I don't know. You know, the [46:19] EHR is dominated by Epic and it's their [46:22] world, you know, and so I've just kind [46:24] of deferred to them and tried to stick [46:27] where I can have an impact because, you [46:28] know, we partner with them. We think [46:30] offering um pricing particularly [46:33] pharmaceutical pricing at the the point [46:36] of um script where the doctor is there [46:38] offering them different options is a [46:41] >> part of that new announcement that they [46:42] just made about the [46:43] >> Yeah, exactly. Exactly. with Epic and [46:45] Amazon etc. Um so I think that's a big [46:48] plus and I think that's needed but in [46:50] terms of their actual business and where [46:51] it's disruptible I'm not really the one [46:53] qualified for that. [46:54] >> Okay. My problem with, by the way, with [46:56] that announcement was that they I love [46:57] that they announced it and then they [46:59] said it's not going to be available till [47:00] May of next year and I was like, why [47:02] take so long? [47:02] >> I had nothing to do with that. [47:04] >> Okay, [47:05] >> that's typical in this industry, right? [47:07] >> Yeah. Do do you guys think that in 10 [47:09] years uh you know healthcare is still [47:12] going to be a fifth or a sixth of [47:14] whatever it is now of GDP and that [47:15] there's kind of this just sort of [47:17] terminal as it sort of terminal state [47:19] where there's just you know way more [47:20] demand than um than sort of you know we [47:23] can supply and just as it gets easier as [47:25] it gets more efficient we're still and [47:26] as GDP grows you know even 5% a year [47:28] hopefully it's at some point our demand [47:30] is still going to just cap at you know [47:32] 1/5if of the economy. [47:33] >> I do. [47:34] >> Yeah. [47:36] But the value that we get out of that [47:38] spend should be [47:39] >> well the value you get out of the people [47:40] will increase. [47:42] >> Yeah. [47:42] >> Because they'll be healthier. Yeah. [47:43] >> So there's a difference when it's health [47:45] when it's sick care, right, as a lot of [47:48] people refer to um versus when it's [47:50] keeping people healthy and they're more [47:52] productive and there's less downtime and [47:54] you know they're they're smarter because [47:56] it's easier to learn etc. Um so I think [47:59] we'll get greater productivity that'll [48:00] more than account for it which we don't [48:02] get now. Z zooming out a bit, you know, [48:04] Mark, we both we have a mutual friend [48:06] who's a very prominent pollster. I won't [48:08] out him here, but he um he he he he [48:11] believes that or he's done some work to [48:12] indicate that um if you ever ran for [48:15] office, he thinks you could win. Now, of [48:17] course, you you you've said you never [48:18] you you'll never run. Um, so but so [48:20] we're just speaking hypothetically now, [48:21] but I if if you were president or if you [48:23] were advising the the the next [48:25] president, I'm I'm I'm curious how how [48:27] you would uh mo, you know, think what is [48:29] the leverage to most address sort of our [48:31] our healthcare challenges and then I [48:32] want to hear your your broader uh sort [48:34] of, you know, policy positions you'd be [48:35] most. [48:36] >> It's really easy. It's the same thing I [48:37] tell states, you know, that I've [48:39] testified for. One, make all contracts [48:42] with anybody who takes taxpayer money [48:44] public. No redactions. because the [48:47] minute that happens, you start to have [48:49] an efficient market and the costs go [48:51] down. Two, make it so that anybody who [48:54] can find a better price for less than [48:56] their current out- of- pocket with their [48:58] plan, make it apply to their deductible. [49:00] Allow it to apply to their deductible. [49:02] Eight states do that at some level [49:04] already. Um, three, for nonprofit [49:07] hospitals, require them to publish their [49:09] entire general ledger down to each [49:11] detailed entry and all their procurement [49:14] um, receipts and and requirements, [49:18] right? So that everybody gets to see it. [49:20] It's our money. We're the taxpayers, [49:22] right? Four, instead of um [49:27] current insurance plans the way we have [49:29] them, 72% of those big carriers um [49:32] revenue comes from taxpayers. They don't [49:35] take risk anymore. They off it. [49:37] Everything's an arbitrage for them. So, [49:39] knowing that, we also have to consider [49:41] that the one debt we never pay off till [49:44] the day we die is our insurance [49:47] premiums. [49:48] Think about that. whether it's Medicare, [49:51] unless you're on Medicaid, right? You [49:53] are paying till the day you expire and [49:57] when you need it, you have to fight for [49:59] it in the current system and probably [50:01] foreseeable systems. to me if we were [50:04] able to create a program where you just [50:06] deposited what would be let's just say [50:09] how much you'd pay for a premium under [50:11] the an ACA silver plan in your state and [50:14] we just went and did a deal with the [50:15] bank and we said okay in Texas for a [50:17] family of five it's $2,100 at your [50:20] income right so you just keep on [50:22] depositing $2,100 and it cruise and you [50:24] own it but it has to be deposited if you [50:27] want to participate in this program and [50:29] then if you have a um medical incident [50:33] or pharmaceutical incident, right? You [50:35] can take that money. If it's not enough, [50:37] we've worked out a deal with the bank [50:40] where two things happen. One, the bank [50:42] will loan you money up until a a point [50:46] where stop-loss kicks in, which means [50:48] we're going to take part of that money [50:50] you deposit and use it for stop-loss [50:52] coverage. and two, the federal [50:54] government or the state is going to [50:56] guarantee that loan to you so that the [50:58] provider is not out and we're still [51:01] going to, you know, deduct it from your [51:02] payroll so that you pay it back. And if [51:05] you know, you're lucky and you never [51:06] have an incident and you get to save all [51:08] that money when you turn 65 and go on [51:10] Medicare, you just get all that money [51:12] and all the acred interest. Because if [51:14] you think about it, [51:16] the federal government will, you know, [51:19] >> like our internet [51:20] >> cover or um guarantee your student [51:24] loans. They'll guarantee SBA loans. [51:28] They'll guarantee home loans or give you [51:30] a down payment, right? [51:32] >> Third world country. [51:33] >> We'll guarantee all these things, even [51:35] agricultural things. But the most [51:37] important thing, you know, covering our [51:39] health care, good luck. you know, you're [51:41] in a bad accident, good luck to you, and [51:43] we'll maybe push it off to the [51:45] hospitals. They have to take you in. So, [51:47] I would create a plan that says you have [51:50] to, you know, you have to um deposit [51:53] this money. The bank will have a [51:55] guarantee just like a mortgage, [51:57] whatever. We'll loan you this money and [51:58] then we'll have federal stop-loss [52:00] programs above some amount, 40,000, [52:04] whatever it may be. Now, all of a [52:06] sudden, bye-bye insurance companies. Now [52:09] all of a sudden you're not fighting back [52:10] because it's your money to spend as you [52:12] choose. Now we'll have to set thresholds [52:14] on where and how you could spend it. So [52:16] you don't just go give it to your cousin [52:18] Billy who said he's going to fix your [52:20] fix your broken toe, right? But within [52:22] those constraints, I think now all of a [52:24] sudden we have a better system. So [52:27] that's what I would do if I was advising [52:29] um in terms of healthcare. [52:30] >> What what about u you know one of your [52:32] gifts is is storytelling and and [52:34] marketing more generally. What about the [52:36] AI industry in terms of you obviously AI [52:38] is you know uniquely unpopular in in in [52:41] in the US if you were advising you know [52:43] open AI if you're advising anthropic on [52:44] how to sort of ch change those numbers [52:47] and then there's the whole sort of data [52:48] center uh sort of saga I'm curious what [52:51] your advice would be to the broader AI [52:52] industry. I've had these conversations [52:54] right with them and I told them they're [52:56] idiots, you know, because what controls [52:59] how every pretty much everybody in this [53:01] country thinks right now? Social media [53:03] algorithms. [53:07] Literally, I was having a conversation [53:08] with a 20 23-year-old friend that I work [53:12] with and we're talking politics. We're [53:13] discussing who we voted for yesterday [53:15] and how and why. And I said, "What about [53:17] your friends, the other 23 24 year [53:19] olds?" and they said that their friends [53:23] didn't even know living in Texas didn't [53:25] even know who Ken Paxton or James Terico [53:27] were [53:29] >> or are right which is insane in my mind [53:32] right insane in this person's mind but [53:34] it reinforces the fact that you get more [53:37] of what you seek via algorithms and if [53:41] you really have the best AI you should [53:44] be able to not only reverse engineer [53:46] those algorithms but put together a [53:48] marketing marketing program because you [53:50] sure tell everybody you can do marketing [53:51] programs through you know your your bots [53:54] your bot or dot or whatever and really [53:57] start marketing there because this [53:59] applies to politics it applies to AI it [54:02] applies to everything in this country [54:05] the divisiveness we all have our unique [54:08] AI feed in the social media platforms we [54:11] use and those platforms don't give a [54:14] [ __ ] what it feeds you within some [54:16] minimal constraints and So everybody [54:19] looks at the world differently. The days [54:22] particularly for younger people of [54:24] thinking, you know, just buy more ads [54:27] and ruin our football games and our [54:29] basketball games and everything's fine. [54:31] You know, maybe older people vote more [54:33] and that's maybe why, but it's useless [54:35] when you're trying to change hearts and [54:37] minds about AI specifically. Even though [54:40] the irony of it all is that the it's the [54:42] AI behind the the algorithms that's [54:45] ruining everything. But as long as [54:47] section 230 is there, we are where we [54:50] are at. And as long as you know, unless [54:52] we take a Australia type approach or [54:54] some states type approach of 16 and [54:56] under, then my what I told them [54:58] specifically, you need to use your AI to [55:03] segment everything and figure out the [55:05] best way to put together marketing, [55:07] promotion, and however else you want to [55:10] do it. Do it like the Russians do it. It [55:12] doesn't matter, right? Because that's [55:14] what's at stake for you. But people, you [55:16] know, data center hate is just a proxy [55:19] for not being able to understand AI. [55:21] >> Yeah. [55:22] >> And not all know all the pieces there. [55:23] And I don't blame people at all. Right. [55:25] And just like employers have been really [55:27] dumb with their healthc care benefits, a [55:29] lot of communities have been really dumb [55:31] with the the deals they've negotiated [55:33] for data centers, you know, and they've [55:35] fallen for the contractual terrorism in [55:37] a lot of a lot of cases. But, you know, [55:40] to me, educating them about social media [55:45] and educa and educating our legislators [55:48] and whoever's president about social [55:50] media because I don't think they truly [55:51] understand the impact that we don't all [55:55] read the same newspaper, the same [55:56] magazines, look at the same websites, [55:58] watch the same news shows. We all have [56:01] our own unique TikTok feed. Your feed is [56:04] not like mine. Like the best way to find [56:07] out about your kids is to just look at [56:09] their social media feed. Like my [56:11] 17-year-old son, it's basketball, [56:13] basketball, basketball. Now I'm starting [56:15] to see a little bit of politics pop up, [56:17] but then hot girl, hot girl, hot girl, [56:18] basketball, basketball, car car. And I'm [56:21] like, okay, [56:22] >> typical 17-year-old guy, right? Um but [56:26] until the the big um frontier models [56:31] understand completely that they have to [56:34] figure out how to become part of [56:35] everybody's feed. [56:38] They're going to be putting out all [56:39] these position papers that are useless. [56:41] >> Yeah. No, I mean Mark, we agree with you [56:43] that what you said earlier about this is [56:44] like the best time to be alive and and [56:46] an entrepreneur. Um and you know, so [56:48] many things that again were intractable [56:50] uh historically are now all of a sudden [56:51] addressable. Is is there an area of the [56:53] market and you meet a lot of [56:54] entrepreneurs and see you know sort of [56:55] people building in all sorts of areas of [56:56] the world. Um are is there like some [56:59] area where you're like why is no one [57:01] building here like that you think is [57:02] still a whites space opportunity for [57:04] entrepreneurs? [57:05] >> Um all of them you know seriously you [57:10] know when I go talk to kids um and I [57:14] always tell them there's going to be two [57:15] types of companies. Those who are great [57:17] at AI and those who have gone out of [57:19] business. And if you take the time to [57:21] learn and then there's two types of [57:23] people and how they use AI. Those who [57:25] use AI to learn anything and those who [57:28] use AI so they don't have to learn [57:30] anything. And if you're one of those [57:32] people who uses AI so that you can learn [57:34] anything and you realize what type of [57:37] companies and problems aren't being [57:38] solved by AI, don't be shy to just talk [57:42] to your AI, you know, and I say this not [57:44] just to kids to 50 year old, 60, 70 year [57:46] old entrepreneurs. The biggest challenge [57:49] right now with people, entrepreneurs [57:52] using AI is the comfort level of talking [57:55] to your AI, whether it's by voice or [57:58] typing, whatever it is, because it's [58:00] there's still a lot of hesitancy. You [58:03] know, just like back in the day, those [58:04] first spreadsheets, people still had [58:05] their 10 key adding machine and they [58:07] would check it to make sure the [58:08] spreadsheets were right. You still you [58:10] do have to check the output but you have [58:12] the opportunity to take the output of [58:14] gro and put it into chat GPT and the [58:17] output of like when I go through all [58:18] these contracts I was telling you about [58:20] I start with one and said hey one of [58:22] your peers came out and said this about [58:24] this contract what do you think and [58:25] spend a couple hours that's all young [58:28] entrepreneurs so when you ask me is [58:30] there one thing I truly believe it's [58:33] everything everything will change this [58:35] is the greatest transitional technology [58:38] we have ever seen and I'm on I've been [58:40] around for all of them, right? And this [58:42] blows them all away. The hardest part is [58:47] just going for it. And that's where [58:48] Silicon Valley and, you know, kids from [58:51] the the Bay Area, I think, have a little [58:53] bit of an advantage because they see [58:54] their parents and friends and the [58:55] culture there as little go. [58:58] But I think the rest of the country and [59:01] the rest of the world once they get that [59:03] confidence level, you're going to start [59:05] to see businesses we never imagined pop [59:07] up because we're not seeing that yet. [59:09] You know, we're not seeing companies [59:11] where it's like, oh, what is that? You [59:13] know, why are they doing that? You know, [59:15] what is the opportunity there? And then [59:17] they realized it before we did, right? [59:20] It's kind of like Fster, you know, first [59:22] like, whoa, what was that? And so I [59:25] think the biggest [59:28] thing to focus on if you're, you know, [59:31] looking for something is ask anything. [59:34] You I mean, I do it all the time. It's [59:37] just like, "What do you think about [59:37] this? What do you think about this? What [59:38] do you think about this?" You know, I've [59:40] held phones up next to each other and [59:41] take the output and had them have a [59:43] conversation with each other so I could [59:45] listen in. Don't be shy about using AI [59:48] in things that feel really uncomfortable [59:50] and weird because that uncomfortableness [59:53] is probably the path to something that [59:55] is groundbreaking and gamechanging. [59:57] >> That's a great note to to wrap on. Mark, [59:59] thank you so much for coming on the [60:00] podcast. [60:00] >> It was fun. Thanks, guys. Thanks, Julie. [60:02] Thanks, Eric.