The Checkbook of the Future: Who Holds the Keys?

Captions as YouTube publishes them, usually machine-generated, so product and people names can be misheard.

1. 0:09

   Um, we're here to talk about the checkbook of the future. It's a timely topic topic and we have a great uh panel of folks here to help you understand it a little bit better. Um, we have Stephanie Zoo who's here from Kaledan uh automated trading firm that does more than $170 billion a year in volume.

2. 0:30

   Uh we have Kingabas who's the COO at MPCH uh infrastructure firm for automated capital and we have Aish Gar who is the managing partner at electric ventures who invests in these protocols and systems that will be using electric checkbooks. So if you are building the digital checkbook of the future right now, I don't see too many of you vibe coding during the panel which is good.

3. 0:55

   We're going to talk about all the features it needs to have and how it needs to work and what it needs to do. Um, so let me start it off with I think maybe the most important question which is trust. Um, what would it take for you to trust a future checkbook?

4. 1:15

   Um, I think we would kind of think about this from a philosophical standpoint first, right? like if you want to fully automate your entire stack um with make sure that all your capital is autonomous um what is the purpose behind that first um when we think about at least from a trading perspective uh you know a junior trader or um someone who is very new to coming into the trading stack that we

5. 1:39

   normally work with um we put a lot of limitations in in the beginning right there's position limits that are hardcoded there's risk managers that look over what they do um they can't move a lot of the uh money off of the platform. Um and so I think you know trust is built up and is gradual. And so working with autonomous agents is the same thing. You're not going to give them all the trust very in the beginning

6. 2:03

   um because there's a lot of edge cases and considerations that you may not have um already. So you need to kind of gradually train and build up to that. Nice. So I'm hearing constraints. Yeah. Ka, do you want to take it? Yeah. So we're at NPC. We're critical infrastructure firm. We're providing security, right? So we don't trust anyone by definition. So from our perspective, we look at it very pragmatically when we're supporting the different kinds of projects around involving a agents and automated like

7. 2:32

   automation in general. From our perspective, we we like to impose governance around these by delegated authorities being curtailed and programmable authority. What I mean by that is we separate different layers of of security around execution versus the escalation framework. So you're not allowing your agent to define the ramifications and envelope they are existing within. So that's how we're looking into that and that creates I

8. 3:03

   would say full trust but it creates an an environment where you're you're separating the different types of authority layers and creates some separation and friction. Great. So uh authorization and authority is going to be at the center of any kind of a checkbook. So do you maybe want to help us understand some of the different nuances in that authorization layer?

9. 3:24

   Right. So we're we're thinking about multiple layers. There's a protocol level, right? That everyone understands what you can do. The second layer is the signature layer which where we are involved with with with with key management and that's cryptographically managed, right? So in the traditional world this already exists, right? I'm not saying anything that you guys don't know. However, it was it's humanor oriented. It's humanbased given that we're giving up that authority and

10. 3:48

    giving the keys the kingdom instead of the kingdom gs we're giving keys to a lower level layer where um the protocol tells you what you can and cannot do. This is your envelope. But then there's a policy layer above that that sits in the signature level where before we were just I would say done signing but signing for based on authority like you you're you're the person who's signing cryptographically. I'm allowing you do stuff right that's not sufficient

11. 4:15

    anymore you need to create the policy and and build that into the signary framework and then you have the third level which is the institutional who sets the policy ultimately it the the future is not about you know who has the keys right the keys will be given to an agent that's just to me it's a fact it's a question who sense sets the boundaries how are the boundaries created and

12. 4:38

    what's the constitution around the framework so what is immutable what is what is the framework that is sitting somewhere cold that it's untouchable and you can't change it. So that friction that three layer of friction at least and we can create more obviously um that will at least give you a minimum viable framework to operate on. So okay Avatar how about you what do you need to see in a system or protocol before you consider deploying real capital? Um, a couple of thoughts. I

13. 5:11

    think um, it depends a lot on the use case. Um, and I I was scribbling notes because I was like I think it's an interesting question because I think there's like one way to think about it which is there's the app layer, there's the info layer, and there's sort of the the protocol layer and they need different things for for things to get hardened. Like at the app layer, you just you have to do it by use case and

14. 5:34

    you want to think about um you know permissioning and who which users have access to what stuff. At the infra layer, I think there's a lot of messiness right now. It's actually an opportunity I think for a lot of engineers to clean up infra like how does the money actually move around.

15. 5:46

    Intents obviously help a lot. There's a lot more to do. A lot of that I think is going to be a combination of like just Lindy and finding bugs and being open source and being formally verified so you can find all these things. And then I think at the protocol and settlement layer um you know I think there's likely multiple layers there and uh some of them will have you know very fast finality for small transactions and and maybe the ability to do some sort of

16. 6:11

    like dispute resolution and then there'll be some base layer that's like immutable and and so on. And so kind of the answer depends on like where you are in the stack in terms of what you need to build and what what we need to build to sort of make these things robust. But I think you kind of want all of those things. And once we have all of those things, then and only then will you be able to say, "Hey, agent, like go do the thing and I trust you to do the thing." Because there there's so many places where it could fall, like it fall apart

17. 6:34

    at the app layer. You gave it too many permissions. It did the wrong thing it wasn't supposed to do. It could be like the money moving around, sort of imploded in some way and went to the wrong person. It could so I think we actually have a long way to go before the stuff is pretty pretty hardened.

18. 6:47

    So Stephanie, you're you got real volume in automated trades every day. What what breaks? What goes wrong? Which of those layers are the the problem ones? Yeah. So, I think from a trading firm standpoint, right, Alos and algorithmic trading in general, um we run, you know, thousands, millions of trades a day without any um human interaction. Um but the thing is is that the per trade bases

19. 7:13

    is very small compared to the entire exposure and there's like a lot of risk management um when it comes to how much is aggregating um across all the different exchanges, right? Um and so in that kind of nonaututonomous but still algorithmic trading there is still a lot of room for error. I think a good example is in like 2012 Night Capital.

20. 7:34

    Um it was a traditional fund um algo trading and they deployed a software that accidentally did something wrong and within like a couple hours lost $450 million and it took them 30 minutes to go in and make the kill switch. And then on top of that there was like 90 notifications that something had gone wrong even after the market opened and nobody did anything right. So it's not just about the fact that um there is like independent decision-m you have to

21. 8:07

    have all the parameters and all the um functions set up so that when something wrong does happen you are able to step in and uh like make an action. Um, and I think that like when the autonomous agent comes and notifies a human, knows when to step in and I think is also very important.

22. 8:26

    I mean, from a from a founders perspective or an engineer's perspective, I think this is this is the critical insight is like you can move money around now. It's technically possible to do the thing. It's like an MVP level. You can build code to do it uh and build apps to do it and build agents to do it, but the hard work is now all in like the risk mitigation and and that's going to take a while to sort out. Um, but if you're if you're the kind of person that likes thinking about those kinds of problems, um, it's like

23. 8:50

    an adversarial thing. What are all the ways things can go wrong? What are all the ways this could blow up? How do we have the right kill switches? It's actually a really interesting opportunity for the right kind of person to want to go build.

24. 8:59

    But I think it's much more difficult in financial markets than it is in like designing cars, right? Because it is like cars are always in the realm of physics and those rules never change versus markets are reflexive. And if an autonomous agent makes a very large trade, it can move the whole market and it changes everything. And then, you know, there's so many variations and teslations of what's going to happen.

25. 9:23

    It's chaotic in a different way. Yeah. Well, so let's talk about accountability. Let's talk about who's on the hook when things go wrong. Uh I think you're right. Night Capital showed us that things can go wrong. But what happens when we get into systems where the AI was operating inside some set policy. There were some key signers that may have been dumb signing transactions and now we have a

26. 9:50

    treasury that's collapsed or some kind of, you know, banned transaction that's occurred. Who how do we think about who's responsible and and what happens? Well, in in that scenario, obviously then the the regulator or whoever is is in charge is not going to go after the protocol, right? And and the signing layer, and we can't just go and hide behind, well, you know, we we'll signed it and we don't sign things. The reality is is that as I said, when you create

27. 10:17

    the envelope around the actions Stephanie said, and night capitalism is is a is a great example with that, right? um what happened tonight post night capital the CEO resigned the company almost went bankrupt right so there are real real consequences of these actions and I think the ultimately if you design a a flawed system a flawed authority system you will the the regulators will find the human behind the the governance right so that's just the fact of life so

28. 10:49

    you may as well as uh as we said earlier is you may as well just design around the governance, right? And and make sure that you're you're playing by the rules to the maximum extent you can. However, with with programmatic approaches, right, full automation and autonomous systems, it's very hard to live by the standard rules we have today. Standard rules we have today is human based, right? So, we have an organization. I come from finance. It's very simple. You have three, four layers of of security.

29. 11:20

    You go and you have a human. It's pretty static, right? You can sign or you cannot sign. It's attached to a person. However, in a in a fully auton autonomous world, a human in the middle is not sufficient anymore. A binary constraint or static constraint is not sufficient anymore. The the system itself has to be able to work with programmatic authority. Not just programs of signature, who owns the key.

30. 11:48

    It is how do we react to key changes. So you're I'm thinking in these systems of not one agent. I'm thinking a collaboration of multiple agents in in this scenario where you you separate the um the execution layer from the escalation layer and then the constitutional layer which is which is immutable. So in this scenario I can see see where the execution agent is is in milliseconds reacting and acting and rebalancing portfolio. Let's talk about stable coins here for a second where we

31. 12:19

    know where agents where agents going to go next, right? If they get authority and then the second layer is we're we're we're taking a look at what's happening in the market today where where are margins going, you know, is there any any event that's changing? So there's a stress in the event. That stress has to be programmable. The the system has to react to that. So we we should not be coding for the standard static limits of

32. 12:45

    you can do 20 million trades, you can do this, you can do that. We should be stress testing our scenarios and making sure that we have a programmatic reaction to that and we have an absolute escalation mechanism by which we can decide and create a friction between the execution and the governance so that we're not blowing up the system. So let's talk about that escalation kind of authority and and whoever that ultimate

33. 13:10

    signer is like you you said the current system is a human-based system current regulatory system. Do you see that changing like are we going to be if we fast forward 10 years will we have a different kind of system or will we have every automated trading entity needs to have a chaperone human chaperone?

34. 13:32

    I think it was weird combination. I don't think it's it's binary. It's not going to be an either or. It's fully human for fully automated. We've seen fully automated systems failing quite a bit. Right. It is about creating the friction be between execution, escalation, and Yeah. And I got it. But ultimately at the end of your escalation path is a person.

35. 13:51

    It's to me it's a person still. I think there always needs to be a person, right? Because there's a you can't punish an AI. I mean ostensibly you can deduct points or money or whatever from them but in the end you know you can you can turn them off technically but they don't care right so in order for you to hold somebody responsible in responsibility consequence and the only like entities you can really punish are human beings or or you know entities either like a legal entity right a

36. 14:20

    company or yeah cut you off I I I was thinking about this actually and I've ranted on this before. So, apologies if anybody's heard this before, but I think um so today we what we have is like a human like one one way to to use a different word for what you're asking Rob would be liability like who ultimately has liability. Um and that's just sort of the other side of accountability. And right now what we

37. 14:44

    have are essentially human controlled and human mediated systems. Um and that really breaks into two different kinds of systems. We have sort of single jurisdictional systems and multi-jurisdictional systems. And and really in finance, we've we've generally tried to keep things simple by having them to be like single jurisdiction. So things exist in like Delaware, right?

38. 15:03

    And so if something goes wrong, you go fight it out in like a Delaware court and there's a bunch of rules when there's a bunch of case law for the last 150 years to figure that out. Multi-jurisdictional starts to get complicated. Like what do you do if it's like a developer in Malaysia running code on some server in the EU and then like something bad happens to like somebody in the US? This starts to get like kind of gnarly already with the internet. And then I think what what we're talking about here what Stephanie

39. 15:27

    was talking about I think you start getting into this like sovereign situation which is a whole brand new world which is like what happens if there's not a human behind it at all. It's like some piece of code that owns a wallet that's executing code to make more money to buy more compute time and it's basically was written by somebody and we don't know who that person is cuz it's open source and it lives on the internet. Uh like how does liability work in that case? Um and I actually don't know if we have we don't I don't

40. 15:50

    think we have a good answer for it. I don't know what the good answer is, but it reminds me of um if anybody has read the history of the limited liability corporation, which sounds super dry, but is actually super interesting. Like, raise your hand if you've read the history of the Limited Liability Corporation. Yeah, there's like one guy. All right, two people.

41. 16:06

    So, it's a great book. Everyone should go read. I think it's called The Company. I forget the author. Um it's been a few years since I read it, but it goes back and it recounts the history of the Limited Liability Corporation. And then like around the industrial revolution around the 1850s, this was a hot debate in the UK and the US which was like what do we do here? Like we have um and and and by the way what I think is really interesting about this is this isn't this isn't like some weirdo edge case legal ease like thing.

42. 16:31

    It's actually a technical breakthrough, right? Because like when you create companies like the first company was the joint stock corporation um with uh the Dutch East India Company in 1648 and it was actually a technical breakthrough because you now had people who could pull capital and pursue risk and then there were people that were like doing the work and you could separate those concerns um and and people who gave in the money could um could benefit from that and all as a as a side effect of

43. 16:55

    that you could pursue tasks that you wouldn't be able to pursue otherwise. So for example like factories and railroads just cost too much money for one person. there's just no one maybe Elon Musk but you know back in the day and so actually the the company was a really important technical breakthrough but from about 1648 to like 1850ish 855 you had to go to the king or you had to go to your state's governor to get a charter to

44. 17:17

    start a corporation um and uh that's a lot of friction um and so people started having this conversation of like well why do I have to go to the king to get permission or why do I have to go to the government to get you know permission and get a charter what if I just filled out some paperwork and I could start a And so then people started having all these debates which is similar to what what I think we're starting to have with AI agents and sovereign AI agents which was well how does liability work right

45. 17:41

    or like how does free speech work like are these things allowed to just say stuff? Um and there are these like really fascinating um debates that people start having about like well what if these things get a lot of money and then they have a lot of power and then they start influencing our political process. Like do we want that? Is that allowed? Um like what happens if a company accidentally kills somebody?

46. 18:01

    like are the people that are behind the company? Do they go to jail? Like what what how does this work? And it took us like 50 or you know years to figure that stuff out. But a as we figured it out, it gave you the industrial revolution.

47. 18:11

    Basically it was like a it was like a parallel technology breakthrough. And I think we're dealing with something like that now with agents which is like we kind of don't know the answers. Like how does free speech work for these things?

48. 18:20

    How does liability work? Is the human who wrote the code liable or is does the free speech protections that we have today mean that you wrote the code and therefore you don't have downstream liability? um you know like how do taxes work? You know, do these things have to pay taxes? Cuz if not, maybe I should like hand my money to this thing and it just does a bunch of trades for me and I

49. 18:36

    don't have to pay taxes and then one day it just like pays me back somehow, right? So anyway, I think we're going to hit like a bunch of really weirdo edge cases here. Um that we haven't had to think about for like 150 years. Um and actually for for anybody who nobody here is you know in Senate or Congress, but I've literally had this conversation with Congress people and senators like I think it's a seminal piece of legislation. Like I think somebody sat down for a year and thought about all these problems and created a framework

50. 19:01

    for it. I think it's like a thing that you could put your name against that would basically go down in history as like oh this is how all of these questions should work. I I think it's a really good question because we don't like nobody I've talked to has the answers to any of these questions. Great.

51. 19:13

    Do you think who's setting the boundaries of the the envelope? Whoever is creating the policies is ultimately responsible. So you're you're you're thinking you're questioning that kind of authority layer that look I created the governance around it. I told you what the constitution was. Yeah. Turns out that was wrong, right?

52. 19:30

    Yeah. Well, I think it it hits on a couple it hits on that. I think it hits on what Stephanie was talking about, which is like, well, what's the consequence? Cuz like in the physical world, we'll send you to jail. Uh, and there's like a real cost. There's like a real cost to you to you don't want to go you don't want to go to jail for 30 years. Um, but an AI agent may not care if you just shut down the data center,

53. 19:47

    right? I mean, they kind of care if you tell them you're going to kill them. Like they might do certain things. It's like kind of interesting to see, but like practically speaking, they may not care. Uh, and for any agent agents listening where I'm not saying to kill like agents, uh, but you know, it's and LLC's can go bankrupt and yeah, and LLC's can go bankrupt and so there's like a real cost and so you lose your So, but like what is the what is the stick in this situation?

54. 20:06

    The stick is the tokens, right? Like you take away your token, you turn off like so there's ways to to demotivate the agents as well, right? Totally. Yeah. Well, and and I think that's where this gets interesting too, which is like that's an economic so as as like you know, investors, you know, we can think in in those terms. What do you do when you have a non-economically motivated actor like a state actor, right?

55. 20:26

    Um, and we always sort of deal with this like we dealt with this at Facebook uh all the time. Rob and I worked together a long time ago there and it's just like yeah, what do you do when you know a state actor basically North Korea says like I'm just willing to spend money because it's not about the the financial outcome here. It's like the the the disruption is the point. Um, and so how do we deal with that? Right? Today we can we can do sanctions on countries and shut them off from like the global financial system or we can shut them off

56. 20:49

    from goods and services. But if they launch a bunch of autonomous code and agents that like goes out there and starts doing stuff, they don't care if they get shut off and like what does it really cost? Like the asymmetry in terms of power projection is so so different here. So anyway, I these are like really really gnarly problems that I don't think we have good answers to. Um but it's back into this umbrella I think of what what Stephanie was talking about which is like it's all risk mitigation right now. Like I think we could see

57. 21:12

    what's happening and it's like the hard problems are now thinking through these edge. Okay. Sure. Go ahead. I was going to say I think that crypto has a much bigger opportunity to solve these problems or at least to build sandboxes in which these problems can be tested out um whether it be like edge cases or trying different types of constraints than traditional finance does right and

58. 21:32

    whether or not you're testing with kind of economic incentives or um you know reputational or other aspects um you know there's I'm sure a lot of people in the audience can kind of play with these different things and build these layers into crypto and because of all the smart contracts s and everything. It's so much more applicable.

59. 21:49

    Yeah. So, there's a I I was going to ask about what what sort of financial control is most scary um if we kind of think forward, but I'm going to change the question. Uh and I actually think the the LLC, you know, um metaphor is a good one or example is a good one. If we could fast forward 50 years from now and think about the kind of what what is going to change in a positive way, what are going to be the kind of amazing benefits of this kind of automatic

60. 22:24

    automated agentic sort of control of money? How will our lives be different and better? Very interesting question. Very good question. Um, and I I take a crack at it, but I want I want to hear your obviously because you're an investor. So, I think from from our perspective, right, um, you can do a lot of good with money and a lot of bad with money. So, and and I think the good thing is that you remove frictions in in in the in in moving money around

61. 22:58

    areas where you know where currently people are on bank for example, right? We have been talking about the on banking, the unbank with crypto. That's our thing, right? We want to make sure that we we reduce the friction to minimal so everyone can have access to capital and can fund projects. But going back to limit lab company, right? I can fund a project anywhere in Africa to create something really meaningful. So that friction with crypto we achieve because you have settlements and you

62. 23:25

    have digital currency. However, the the movement of that is is still very inefficient. It's very expensive. It's costly still. So when we're what what we're trying to build with with in my view and the good thing is when we're going in and AI agents will make that so it's frictionless to the point where it's going to be cheap so people can do great things. The bad thing is when we look at the projects emerging today,

63. 23:50

    what are people looking for? Short-term rent seeking, right? When you're saying fully automated trading systems, what are they doing? They're just doing millisecond transactions to extract as much value as possible and create alpha, right? So, you do have this dichotomy and and I think we in the room which I said, right, we need to sit down and figure out where we want to go with this because you can do a lot of good things

64. 24:12

    with this and a lot of bad things. that the question is really not about can you do it and what should you be doing it's happening right make sure that you put guardies around it that we we're moving into the direction that's positive I would say it's not just the rent seeeking behavior right when HFTs were invented in tradi all of a sudden like markets just became so much more liquid there was so much more price discovery you know HFTs are around you know I I

65. 24:37

    work for one obviously but um they're around because they actually help move the markets um and increase liquidity and all these necessary things. And so ostensibly, you know, having agents and agents that are able to react even faster, make, you know, better decisions, whatever that might be, um, you just you have greater liquidity and better financial systems, right?

66. 25:00

    Michelle, um, I'll be a little abstract in my answer, but I my interpretation of your question, it sparked a bunch of ideas. So I think the interesting thing here, one way to think about this is like you have agents which are basically going to be super intelligent um uh creators starting in the digital realm eventually maybe with robotics in the physical realm. They can interface with the physical world through humans in all sorts of interesting ways. Um, and what the intersection with capital lets you

67. 25:29

    do in crypto and stable coins and all these onchain infrastructure to move money is it allows all of those super intelligent uh entities to actually be able to have a capitalist uh framework and have a scaffolding through which they can interface with each other. And I tend to think, no surprise as a venture capitalist, I'm not a venture socialist that capitalism is pretty awesome as a technology. uh you know the ability to to allocate resources towards

68. 25:58

    the things that that create value um and do it much more efficiently than any other system. Um, and so I think what you get with a combination of these two things is the ability to have just dramatic acceleration in taking resources and applying intelligence to them which really creates value and then really supercharging that at at sort of like a systems level with the effect of allocation of capital which capitalism

69. 26:23

    has done really really phenomenally. Um, and so I think you can essentially create like a capitalist economy around these things, which I think is is ultimately the best way to figure out what problems to even solve. Um, and then I think when you think about it from like a human perspective, the fact that now the cost of participating in the economy has come down so far. I mean, you look at per, you know, dollars, uh, tokens per dollar, you

70. 26:48

    know, going through the roof and and the cost of silicon coming down and all these kinds of things. You're not talking about the ability for anybody in the world with in the grand scheme of things relative little little money, right? Like a couple of thousand dollars to be able to take their ingenuity and their creativity and and create value.

71. 27:08

    Um I think is transformative. Like this was this was for people who have been around for a while. I think this was the original promise of the internet. It's like you can just create and put it out there and and your thousand true fans will find you. if anybody knows that Kevin Kelly article from 20 years ago.

72. 27:22

    Um, but it was missing this extra component of of the capital which is like even if I'm a brilliant person, like how does the money get to me? Like I'm some person sitting in Asia or Africa or Latin America and I'm brilliant, I'm thoughtful and I'm smart and I can create and I have the skills but like if I can't get that capital injection to actually get the flywheel going and then scale it up, it's you can't really get off the ground. And I think this solves that. Um and so you have this like massive lever for people

73. 27:47

    uh coming in the form of agentic tooling and code and then the sort of super power that comes with capitalism and the ability for capital to flow freely. Um so to me I think there's there's probably the mental model I always have is like we have one Elon Musk, we have one Zuck, we have you know one Bezos, like pick your favorite um you know outlier. Um, but I tend to think there

74. 28:09

    probably 10 to 100, you know, like one to two orders more magnitude such people out there in the world just probabilistically if there's seven billion people. Um, I don't think we've actually tapped more than 700 million of them. And so that's extremely optimistic. I think it's extremely bullish that there's actually we could have 10 SpaceXs and we could have, you know, you know, 100x the robotics companies. We could have 100x the, you know, imotherapy companies to cure cancer. And like the the rate limiter

75. 28:32

    has just been intelligence and capital and and now we kind of have like solved both. Nice. very optimistic vision. I love it. Um, okay. So, let's leave people, we got a minute left. Let's leave people with a spot to watch. If your vision sort of comes true and and we start to see this kind of, I don't know, explosion of, let's call it, capitalistic creativity and ingenuity, what's something that hasn't happened yet that will happen that will tell us that we're on the path?

76. 29:03

    Um I think um well we we already are seeing some of these. I mean I think you're going to see the rise of like one to five person companies that have global impact. So you look at something like OpenClaw and you're just like wow this is amazing right? Like I can't believe this thing took off and the internet supercharged it and it's open source code. And so I think you're going to just see a bunch of those. Um and that's a signal that something is really

77. 29:25

    happening. Um, and then I think for me the thing that's going to tell us that we're not just at that sort of early stage, but like it's really starting to scale and have impact is highly regulated industries that have historically been very um slow to change all of a sudden will start to get disrupted and start to change. So for example, education and healthcare um where people because they'll have super intelligence will just circumvent the

78. 29:48

    whole system. And so now it's like well why should I even go to school if I can just get like a one-on-one tutor that's way better? And I think somewhere in like the next five years we'll we'll tip on that and then everybody just gets like their own personal super intelligent tutor or their own personal super intelligent doctor. And I think we're like only three to five years away from most of that stuff just kind of tipping. Nice. Cool.

79. 30:08

    Okay, we are out of time. I'm going to give either of you the final word if you want it. No, I I I just wanted to say I totally agree with but we can see is rapid prototyping solutions to big problems. So even if you look at diagnostics you know in healthcare right that's just moving so much faster and that's what we're seeing already happening right so what products took two years three years four years five years it's taking a week like that's the that's the rapid improvement

80. 30:34

    great way to end it it's in your hands it's in your hands exactly thanks everybody