Viktor
00:00:00.000
It's a good analogy, right? Just as serverless is misleading, kind of, oh, there are no servers, right? It's rather that we don't see servers that much ourselves, That's the thing with serverless. Servers are equally there, and probably the same thing with zero human companies. It's a very misleading name, but if, the point is we can do so much more with fewer people, then I'm on board This is DevOps Paradox, episode number 367 Zero human companies still need humans
Darin
00:01:35.069
we're a founder, right? We have to be the founder to actually get that company going.
Viktor
00:01:40.001
Oh, that's cheating. That's cheating. That's not zero human company. That's one human company
Darin
00:01:48.506
but we also need a human when the AI CMO hallucinates some new marketing material, so the headcount isn't zero, but we still have to have at least one human. So we're thinking about this. I've been seeing people pitching this about, "Okay, we're gonna have nothing but an AI-led company." It's like, uh, okay uh, no. Especially from the VCs. They're talking about hallucinating. They're really hallucinating
Viktor
00:02:19.354
if the story would be there is going to be one expert in something and another expert in something else instead of an army of experts in that something, then I would say probably, yeah, that sounds doable. Sounds interesting. But, I know in what I do that there is no way that my daughter would do equally good job with AI. There is absolutely no world in which that would happen today And I cannot imagine that if we would invert those roles and say, "Okay, Viktor, now you do marketing campaign," that I would do as good of a job as somebody who is really good at that something. Now, I would do much better job with AI than I would do without, but it would still be bad, just slightly less bad
Darin
00:03:09.554
I think the framing of this too is the humans haven't left the company. It's not like Elvis. They haven't left the building. But they've changed roles. And in fact, you had a video over on your DevOps Toolkit channel that came out on July 6th. I don't remember the name of it, but basically it said, "Hey, I'm now the CEO of a company." And that's a possibility now. That was not a possibility a year ago
Viktor
00:03:33.785
Yes, that's a possibility. I still am skeptical if that would be a real company in that video that does everything the companies do it would not work, I mean, I say I'm CEO yeah, I can work alone on my projects, but I'm not doing real marketing, I'm not doing sales, I'm not doing a bunch of other things. It's open source, I don't care. I mean, I do care, they're my babies, but not to that level. If it would be a real company, then uh, yeah, I would need people, let's say maybe 10, What I'm equally… So I'm, I'm very skeptical of that zero human or one human, I'm the only person and there is an army of AI agents. I'm not sure about that. But equally, I am confused as to what role will the current massive companies have in this world. let's say Microsoft, and I'm not picking them for any particular reason. how many people Microsoft has, like 100,000?
Viktor
00:04:32.755
let's say it's 100,000. Uh, You don't hold me that-- for that figure as being accurate, right? Do they need that many people? We can discuss what is the correct number, but it's not that It's definitely not that. Hundred thousand? Not hundred thousand
Darin
00:04:50.165
That's possible, especially if they are able to do that. Let me give you a real-life example on this. This was from a client that I was at a number of years ago. It was an insurance company, and people keep paying premiums into insurance companies, right? You get used to that. And somebody told me there, and I'm not gonna name them, but somebody told me there, it's "Hey, look, even if we did not sell any more policies starting tomorrow, we could get rid of everybody except for about 10 to 20 people to write the claim checks when there is a claim put out against the policy." Think life insurance, think, you know, whatever, right? 10 to 20 people. That's like our company would still be solvent for another 250 to 300 years without issue
Viktor
00:05:35.611
You know what's the problem that company has? Maybe that's a whole countrywide problem. Writing checks
Viktor
00:05:50.705
Yeah, exactly. Exactly. I mean, I, I don't know what the numbers are but they must be lower. here's what I think it's a paradox, is that I don't think that they will have higher unemployment I just think that we will have larger number of smaller companies than we have today Smaller comparatively to what we have today, whatever smaller means
Darin
00:06:13.485
I think there will still be the hypers of whatever those are. You're gonna need those in order for
Darin
00:06:21.535
hyperscalers or hy- hypersized companies. there will still be a handful, but we're not building… In fact, you don't see that a whole lot. car company Toyotas of the world or the GMs of the world or you could call it SpaceX, whatever. There's still gonna be large companies in a manufacturing perspective. You're gonna have large companies that provide compute
Darin
00:06:44.097
for all the small companies. But beyond that, do we really need a large company? I don't know
Viktor
00:06:50.742
even those currently large companies, should they stay equally large in terms of headcount? That's I feel the question. But because if they stay the same in terms of headcount, and let's say that the world is moving or already moved towards being-- It will definitely move towards being much more productive, right? So if the headcount of those companies stays the same, does that mean that they grew like, exponentially in terms of volume of what they do? If they don't and the headcount stays the same, then they're losing competitive edge So because if everybody's more productive you either do more with the same number of people or you do or you do the same with fewer people Or you're not more productive. You can pick any of those three options
Darin
00:07:38.832
The second one is interesting because if you do the same with fewer people, that's good. Let's say that's good. But then is your company growing? You may not want to grow your company beyond a certain point,
Darin
00:07:53.000
so therefore you don't need a lot of people to do that. It's the first one that's a little bit harder
Viktor
00:08:03.898
then you cannot not grow. Private company, r- let's say that I'm a private company. I have a private company, And I'm earning good money, like good money for me, right? No, not Jeff Bezos good money, but good money. I can say I, I don't wanna grow. This is fine. This is great, like I'm set for life. I'm good. But if you're a public company, you need to have growth. Otherwise, your shares go down
Viktor
00:08:28.860
Yeah, exactly. So you cannot not grow, and that, that's maybe the problem with public companies, Actually not the problem. It's probably good that everything grows, but that's a constraint. You cannot not grow. Let's say Microsoft, just to keep the same companies before, what was it? Earlier this year, their stocks started falling even though they had tremendously good growth, like double digits and that still wasn't enough, and their s- their shares are really plummeting, so they cannot not do that. They cannot not grow
Viktor
00:09:10.858
That, that's a conundrum, right? Like you cannot be agile enough when you're that massive. ask anybody, let's say ask a Googler how easy it is to bring idea to life, and you will see that they're as administrative as it can get. And this is not me trashing Google now simply because maybe there is no other option in, at that size
Darin
00:09:31.486
Let's flip it around the other way. There is an app, for lack of a better term, called Paperclip, where you can set up a company yourself. So you're the human, you're the CEO, you're the whatever, and you start defining out, "Okay, here are all the roles I need for my company." And then it just goes off and does it, as long as you have reasonable amount of money to pay for all the tokens you're going to be spending
Darin
00:09:58.446
It should be cheaper than people. Done correctly, it should be cheaper than people
Darin
00:10:06.015
yes Yeah, I don't know how you could do it for people. But see, that's also a problem to where, okay, let's say I'm US-based, I'm gonna start this company but I'm gonna do everything in either the Philippines or Argentina, which typically from the US perspective have been the cheaper companies to pay people in
Darin
00:10:26.341
I don't know that, that it would be cheaper than that, but it's also harder if you've never set up and run a remote company before of all outsourced, for lack of a better term, people, then what's the chances of your company growing versus just spending maybe a little bit more and having AI do it all for you?
Viktor
00:10:49.725
But, you know, outsourcing will change drastically as well because, you know, outsourcing in terms of looking for cheaper labor, not outsourcing in terms like we are looking for an expert that is not in our village. But when it's economical decision that decision makes sense when labor cost is significant percentage of the total cost, and now it's not. ' Cause a person in Philippines, in Spain, in the US is going to sa- pay more or less the same for tokens and other things, and when that becomes big part of the expense, then, hey, whether I'm hiring for 100K or 50K or for a person does not actually makes that much of a difference anymore if that's only a small part of the cost
Viktor
00:11:43.699
what I'm rather trying to say is that this might result in, okay I'm not outsourcing for the price. I'm outsourcing for the experience because the price is, does not differ anymore that much because it's becoming smaller part of the cost
Darin
00:11:59.128
let's talk about this as a potential analogy that's going to fall apart really fast. Uh, is, Is zero human the new serverless?
Viktor
00:12:07.645
It's a good analogy, right? Just as serverless is misleading, kind of, oh, there are no servers, right? It's rather that we don't see servers that much ourselves, That's the thing with serverless. Servers are equally there, and probably the same thing with zero human companies. It's a very misleading name, but if, the point is we can do so much more with fewer people, then I'm on board
Darin
00:12:32.397
What about this? What if it's m- more like the paperless office or the Linux desktop year? It's always one year away,
Viktor
00:12:39.509
Hey, if it's like Linux desktop situation, then we are in big trouble given that significant part of everybody's economy now depends on AI. That turns out to have the same future as Linux desktop, we are doomed
Darin
00:12:54.444
if I'm a company owner today and I'm thinking, "Okay, AI's gonna solve all my problems so I can get rid of most of my people," I don't know that you're gonna be able to do that
Viktor
00:13:03.950
Look, if you're that person, I have very good advice for you change your profession to grow tomatoes or something like that. You will be more successful at that endeavor
Darin
00:13:22.300
I don't know how you're going to be able to solve… if you're a services business, most people have… Let's put it this way. In the services business, people use a lot more common sense, the plumbers, the HVAC, the electricians than I see in the technical space, like in what you and I work in. They use a lot more common sense. They realize that I have to go to a job site and I have to actually have guys dig a hole. AI's not going to fix that
Darin
00:13:55.985
I don't know that I'm pointing at it. I'm just calling it out blankly what it is. we think that technology can solve all of our problems when all we need is a shovel, a literal shovel
Viktor
00:14:06.794
the truth is in between all this, right? Technology does solve problems. Not necessarily all technology, but in general, technology solves our problems. There is no doubt about that. If you think that there is doubt about that, then I think you should be living naked in a cave without a knife or a lighter to light fire, right? That's all technology. Technology solves our problems. It's just that we have a tendency to exaggerate the scope of the problems something solves, What did you call that office full of AI doing everything for you, right? That does not work. Will it work in the future? I don't know. It doesn't work today
Viktor
00:14:46.769
It's possible, yeah, but I don't know the future. Whomever says that they know more than a year in advance that they're fooling you. If you believe people who say that they know, we should already have flying cars and uh, be on Mars, and AI would be doing absolutely everything for us, right? That's simply not the case. Nobody knows the future. But right now, that office kind of, "Oh, I'm going to go to this service. It's going to set up all the roles for me," that simply does not work. At least it does not work if that business of yours has any competition, because your competition is going to have people who know what they're doing, and they're going to be aug-augmented greatly with AI
Darin
00:15:27.375
that's what I was getting ready to say is routine work has been s- if it can be, been sh- offloaded to the AI agents. In fact, your video, again, I'll go ahead and call that back, the way you s- without… People need to go watch it because it's not that long. For you it wasn't a long video. Basically, you're awake during the day, you're checking everything that happened overnight. This is the classical we've offshored to India, Philippines, whatever. We wake up in the States, look at it the next morning, work on it all day, and then yep, everything's good, and submit it, and then you go back to bed and somebody else works on it overnight. The model is the same.
Darin
00:16:07.904
That's what we want to be doing. That's always been like the dream. And with offshoring to other places has always been a crapshoot at best if you're going to get back. Because in order for those… But it's funny because this takes us back to how it really works. In the old days, you would write up a spec, you would hand it off, and they would implement exactly what was on that spec. How is that any different than what the AI agents are doing today?
Viktor
00:16:35.894
There is a massive difference. Yes, I think that there is a massive difference, and that difference is in timeline or the duration of something, Because, hey, I give a spec and then you come back to me half a year later, and let's see whether that spec makes sense or no when it faces reality, that was the model. Now it's kind of, okay, oh, I'm so sleepy, let me tell you what to do. And I wake up and it's done, and maybe it's crap, maybe it's amazing, maybe it's something in between. But I haven't lost time I can do over and do over and do over. and let's say it's 50% right. Okay, let's do the other 50 better, right? Let's do it all over from the scratch. We learned something, and so on and so forth. Those are all the things that we could have been doing before with agents. It's just that we don't have unlimited time. It's not kind of, "Okay, so this is completely wrong. Let's do another half a year. Oh, now it's 50% right. Let's do another four months," and so on and so forth, right? You, You end up spending years. So I'm not saying that agents do better job. I'm just saying that their speed makes the difference
Darin
00:17:50.184
So the speed and years, I want to talk about those two things. If I think about it in the old school way, we have, again, it could take months for things to actually turn around and you finally get something, whereas in your case, you can have it the next morning. Do you think part of the problem is that we're, when we're trying to set up our new agents today, are we over-investing in trying to get the evals and the prompts and everything right versus just deciding when should j- the agent just stop and let me get… Are, are we just over-engineering like we always do at the, on the front side, or are we hitting a sweet spot in the middle?
Viktor
00:18:27.661
I guess that depends on a person driving it, It's curious how similar all those things are, right? You can apply any lesson from people before AI, right? Hey whether somebody did a crappy job or no depends a lot on the management and the processes in that company and the culture and all that stuff, right? People giving instructions to follow something through a letter will follow something through a letter, right? People giving instruction to have certain freedom will have certain freedom. Depends. That's the answer. It depends. The favorite one
Darin
00:18:59.402
We're seeing this whole cottage industry also pop up around cleaning up what that side of the fence calls the AI crap code that's been created. My phrase, not theirs, but that's their effect. It's like, "Man, AI's created lots of code for you. That's great, but it's not working. We'll come in and fix it for you."
Darin
00:19:20.772
It would be funny if they said that, but yeah. You know at some point it will come to that
Viktor
00:19:26.822
So what is a crap code, crappy code, right? Is it a code that doesn't do what it's supposed to do? Then yeah, that's definitely without doubt wrong, right? You need to fix it, start over. This is not what I asked for. The-- I did not explain it well, and so on and so forth, right? But there are cra- there, there are levels of crappy code that you can qualify as if this is not maintainable code. I don't care about that anymore, a long time ago, I cared whether it's camel case or Pascal case. I cared whether there are comments or no, but that's because I was doing it. When I was a lead, that was a very unfortunate experience, but when I led people and a project and what so not, at that time, I did not really deal with, "Hey, do you write camel case or Pascal case for functions?" That wasn't my job, and I honestly did not, could not care less. And that latter is the example, Hey, you wrote this code." "Yeah." "Do you understand what that code does? Do you know how to uh, work on it?" "Yes." "Go, Go for it." Just go for it. Same with people as with agents
Viktor
00:20:39.977
whether it's performant. So let's put it this way. What I really care that it passes all the First and foremost, all the black box tests. Oh, I said that there should be a button, and when you click that button, this should happen," is that happening? Yes. Is the performance of that happening correct? Yes. I'm fine. It's good. I don't know whether you wrote 10 lines of code or for that or 1,000, right? If there is no performance issue and if it does what it's supposed to be doing That's it. I don't care about anything else anymore. to be honest, I did not… That, again, that has nothing to do with agents Name me one case when you had… You worked or you were a manager and you said "Okay, now let's see all the lines of code that you wrote."
Darin
00:21:41.968
And move on. I guess really the goal isn't to eliminate… I mean, for some places it will be truly eliminating humans because you've got too many. That's… we saw that to where there was big over-hiring during the pandemic, then things got cut back to where it's a more normal thing. People… It was funny, they were doing talent hoarding. That's the phrasing that I was hearing. That, okay, these people may not be available when we really need them, so let's go ahead and get them, and then they let them go months, a year later. Uh, If you sat around for a year doing nothing, getting paid, I guess that's a probably pretty good deal. But some other people were waiting to try to get jobs during that time. But really what we're seeing is, okay, can we reallocate the humans and move them up the stack? Okay, they won't become the CEO, but as both you and I have talked about, it's like I'm a manager now. I'm a product owner. That's where we want people to be because coding is a so what thing now. Understanding what to do and what should be done, that's very important. But just the act of coding doesn't matter anymore
Viktor
00:22:50.789
No, if act of coding somehow helps you to fulfill those tasks you mentioned, go for it whatever helps you to figure out what should be done and how should be done and so on and so forth is fine. But coding for the sake of writing code, lines of code, bits and bytes no. And this is probably the part where most of our listeners drop off or writing the comments, right?
Darin
00:23:18.119
Because they don't need to write code anymore, so therefore what is their purpose in life?
Viktor
00:23:23.451
Yeah. What's the meaning of life if there is no need for you to write code anymore? And I know at least from my comments I'm receiving that's a very unpopular statement. Very
Darin
00:23:33.704
Why would you want to continue writing code when something can actually generate reasonable, I'm not saying great code, but I am saying reasonable code for you in a much faster way
Viktor
00:23:46.934
first of all, I think it writes great code given time constraints. I dare you to write better code than AI in what it does in half an hour. I dare you if you have all the time in the world, you can write better code, yes. But there is no way that in one afternoon you will do even 5% of what AI does
Darin
00:24:08.548
I will give a story here. I saw a guy this weekend at the time of recording, Fable was just re-released, and we are, we're actually recording this on July 7th, so this is the last day that Fable is available within the subscriptions.
Darin
00:24:25.474
And he had a plan once Fable opened back up to write this native Mac app it is a non-trivial Mac application. And he should be done today with it, and he started on the day it opened back up By the way, he's not a Mac developer, let's put it that way. In fact, he's not a developer at all. But he's able to drive Fable to give him the app that he wants
Viktor
00:24:55.067
And if the app does what it's supposed to do why would he care about the code? Code is means to an end, not the end by itself
Darin
00:25:08.648
Because, so what if you create a great code? So what? Here I am, I'm 40 years into this, and of course that's what I've built my whole career on. Could I have done it differently before now? Okay, sure, we've had some code generation stuff in the past, but nothing like what we've had with AI
Viktor
00:25:26.178
Ah, Darin Close. I never bought the idea of that code generation. I th- I think it was always always ended up with more work than without it. But this is different. This is very different
Darin
00:25:41.419
So what are the roles that we, you know, we talked about being a product owner, but let's say, "Man, I don't want to be a product owner. I don't want to know all that crap." platform engineer is still going to need to be around for now.
Viktor
00:25:54.252
All the roles stay. They might be transformed, but I think that they stay. We need platform engineers, of course we do. We need the infrastructure people, of course we do. We need the applications developers, of course we do and what so not. We need all those people. It's just that you are now being promoted to the lead of the team. That's the difference, Okay, so you were one of the 10 people in the team, there was a lead manager, call it whatever you want, on top of 10 of you. Now you're that person now
Darin
00:26:26.039
Even though the other 10 are still there too, or maybe not all 10 are there, n- maybe there's six there now
Viktor
00:26:31.611
Yeah. now that I mentioned that lead or manager, the best managers are people who have, who've been in working on those things. They know it inside out. Those are the good managers. They need to have additional skills that not every coder, let's say, has, but those are the really good managers, right? People who've been in the trenches
Darin
00:26:55.898
And this, again, I'm gonna refer back to the video that you had come out on July 6th. You showed the basic levels, right? You had CEO, you had managers, and then you had all the different agents doing the things. as a business owner or as a department head, if I can turn 10 people that were doing things into… ' Cause some will probably leave, some might… It's "I'm not gonna do this anymore," down to six or seven, that follows y- the model you talked about. Okay, those seven are basically gonna be managers of a team of agents. Yeah, we've got to wa- watch our budget for token spend and everything else, but okay, great. But think about this. If you're in charge of a platform, right? You're director of platform engineering, and you've got six people and you're getting in requests left right, and center, which is what happens today. as me as an application developer, I am waiting, or as the business owner for a new application going in, I've got to wait on you to put new things in. But now I don't have to wait potentially as long, because it could be there tomorrow
Viktor
00:28:01.248
Yes. Uh, You don't have to wait long, actually what is long depends on how many things you're doing in parallel, So waiting f- if I work on only one thing, waiting for an hour or, or a day for that thing, that feature to finish is too long. If I work on one thing, I need instant feedback. That's ideal, right? But if I work on 10 things in parallel, then as long as one of those 10 is finished, and another one is finished after I'm finished with the one that was previously finished that's ideal, So as, as long as I have things to do, I'm fine. I'm productive. And when I say things to do, I mean productive things, not like meetings, right?
Darin
00:28:45.640
Could you eliminate all meetings? You're going to be meeting with your agents all the time because you're reviewing their work. I mean, let's put it this way. It's a meeting because you're… It's exactly what it is. So how do we eliminate those meetings? Do we create our own AI managers of the AI agents, so then we just talk to the manager, so we're building out our own little kingdom?
Viktor
00:29:05.450
That's the next step, You move up the pyramid. Same with or without agents. You have the lowest level, whatever those roles are. You have then anything between that lowest level and the top of the pyramid, right? The taller the pyramid is, the more you can move up, It's just that you start with a relatively small pyramid. There are only two levels. There, there is you and me, right? Or not even two levels, one level. And then as you master how to manage five people you can say, "Okay, now I will be able to manage five people who manage five people each." And then you can say, and then you can continue moving up until you reach the top. I don't think that we will anytime soon get to that picture where I'm alone at the top, there are no other people, and there are 27 layers below me. that I'm very skeptical in the near future. But if there are more than a few layers, then we need, more than me, let's say
Darin
00:30:07.274
So what you're saying is you would be very skeptical of a four-person company to where you have a CEO and three people, and those three people each manage 100 agents. You would be very skeptical of that. Because of scale, right? Because how would you check the work of 100 agents?
Viktor
00:30:26.934
that's the thing, right? at least in my case, when I let the agents work overnight, I actually reduce the number of things I instruct them to do overnight. Because what is happening right now is that I have things sitting right now while we're talking already for days sitting. There are peer pull requests. I haven't merged them because I haven't had time to go through it. And when I say go through it I'm not necessarily saying I need a day for each, right? But let's say I need an hour for each So if I need an hour for each, when I say an hour, I mean, okay, half an hour-- Let's say half an hour to actually check whether that something works, and then another half an hour to course-correct it, because you're gonna check it multiple times on average. Let's say an hour. So if it's an hour f- to check each of the features, how many can I do in parallel? Eight? And that's assuming that I have no other things to do. I have no emails to answer. I have no meetings to attend, meetings with humans, so on and so forth. So when you add that to the mix, say that I can do a handful of things overnight. I can let the agents do a handfu-- I ca- I can let them do a thousand things at night, but there is only a handful that makes sense to do myself. And let-- So let's say that five features per night, and I'm being very optimistic now that I can review, approve, what so not. What does that mean? If I want hundred things every day, I need nineteen more people and I'm not-- I was never a zero human company. I was one-person company, but if I want to do more, I need to be more than one-person company
Darin
00:32:11.244
Or you make the decisions like, "Okay, I've hit my capacity," so it comes out when it comes out, when I get to it
Viktor
00:32:18.370
what I mentioned earlier about, hey, you can you can do more with less or you can do the same with less or whatever. The point is that you need to think about, hey, this-- there is a limit to how many things I can-- a person can do with agents. And you can say, "Hey that's enough. I'm kind of-- I'm flying solo." Cool. That's the limitation. Five things a day. Or you can say, "No, I want to grow business." Then you need more people. And the moment you get more people, then you need more people for more people. That's unfortunate reality, right? I-If I hire five people, that's… And they manage their own agents and I manage them that's cool. If I hire 10 people, then I need 12 actually people, because I need two people to manage five people each, And then you keep entering that trap of large corporations, right? The more people you need, the higher the percentage of people managing people.
Darin
00:33:14.975
Then I'm thinking on the consumer side of whatever you're producing that I'm using, I don't want 100 things coming at me every day. I can't accept that. I can't accept that risk. Or let's say I update things quarterly, so that means you're putting out 500 things a week. So that, over a quarter, that would be 60,000 things coming at me
Viktor
00:33:36.749
That, that's a very high number. But yes and no, right? I feel that you're looking at new user-facing features, kind of new buttons but there can be invisible features like improvements of performance fixing bugs, for example. "Hey, we c- we can fix our bugs now," right? There are many invisible things you can do that are not really necessarily visible to majority of your users.
Darin
00:34:01.795
Well, you think they're not until the bug wasn't fixed correctly and then causes a whole new set of problems, you didn't think about the chain of the 60-- Okay, not 60,000. Let's say 500. Let's say I upgrade every week. 500 things in there. The chances of something colliding in those 500 is pretty high, I think. Or it's non-zero, let's put it that way.
Viktor
00:34:22.278
this is where I differ maybe with some people, is that the chance of bad things happening, that should not be the worry simply because it's unavoidable. Doesn't matter. You use AI, you don't use AI, it's unavoidable. The question is always how fast, what is the time to detect the issues, and what is the time to fix issues? The question is not whether you have issues. You do. I guarantee that So yeah, if I can turn multiple things every day and some of them end up having an i- having issues like everything does the question is how fast I can fix those issues
Darin
00:35:00.105
But then it's still the same problem again. You're just… We keep kicking the can down the road.
Viktor
00:35:05.800
You're now s- y- you could say the same thing before Industrial Revolution kind of like, "Hey, now you can you, you were able to produce uh, 10 apples a day and now you can pro- produce 100. It's the same thing." No. It's 100 now
Viktor
00:35:22.906
Yeah. Or it's the same amount with fewer people, I know so many startups that grew to s- still small, but like couple of hundred people and say, " You can do it with 20." Let's have 10 startups instead of one
Darin
00:35:41.112
why would we want to do that? I want to have a company that's going to be able to give me a good exit so I don't have to work anymore
Viktor
00:35:51.242
I don't know what to do. I'm not sure what exit-- I mean, w-what exit looks like is billions, What growth looks like is, what was the saying among investors? Double, triple double, or something like that, Something along th-those lines. Th-th-that's… I-if you're still that successful, you're not getting money anymore. Double does not work anymore. It's like double digit, triple digit, triple digit , Darin: quadruple, quadruple whatever the nine one is. Yeah. That, That's what we expect now, or that's what the investors expect. Yeah
Darin
00:36:26.087
Oof. I'm so glad I'm not looking for VC money anymore. just sounds like a very non-fun time to… At least in the tech space. All right, there's still places where VC is normal. Okay, not normal. VC's never normal. But where VC hasn't gone to that kind of chaotic numbering. But that's in boring places. And do people want to work in boring places?
Viktor
00:36:53.100
Oh, they do. So many of my friends when we talk about our jobs, it's kind of like, "Cannot do that. Why would I want to do your job? I go to my job at 9:00, I finish at 5:00 or 3:00 or whatever the time is, and my job does not exist for me anymore. I don't live to work." That's their response. I work to live. But for many people, it's kind of, "This is a job. I ju- I just want to do the job. I just want to get over it." And th- that's boring. Boring is not bad It's bad for me, let's put it that way. Not for everyone
Darin
00:37:27.607
So what's the chances of a true zero human company existing? In theory, there's already one out there, but it's owned by another company, so does that count? I don't think so
Viktor
00:37:37.757
Literally zero human company. Let's say that you get to the point where actually you define one one day everything and from there on it's all autonomous, all agent stuff like that. You at least need a, need, need a person to cash those checks
Viktor
00:37:55.925
Or, at least-- no, just go- If that no human company is earning some money, you want to be at least that one person that, that gets that money I'm assuming you're not gonna make company the, of agents that exists only so that it can finance agents. You wanna live nice as well
Darin
00:38:13.055
Yeah, that's the one I'm still trying to get my head around because not because it's hard, it's just if it were true, I'm thinking especially in the US and how the tax code works here and setting up corporate structures, it's like, okay, if I had set up, let's say I was using OpenClaw or Hermes, I was like, "Hey, go… Here's 1,000 bucks. Go create a, company and you run it yourself and this is how you stay funded. You don't get any more money from me. You get a $1,000 seed."
Darin
00:38:48.622
But then at the s- at some point it's like, okay, oh wait, the thing I told it to build is now worth, let's call it $20 billion. If there's no human to take it out, then who gets the money?
Viktor
00:39:01.378
That's why I want one-person company, not zero people company. I wanna be cashing those checks. That's my job. That would be ideal job. My only job
Viktor
00:39:12.333
is to transfer money from company accounts to my account. That would be ideal. I would love that. Can we, can-- Is anybody hiring for that position?
Darin
00:39:22.678
Okay, so we have to get the transcript right so when the AI agents read it, then they'll, they could come look and send you an email to let you know that, "Yeah, hey, we're actually looking for a human to send money to. Would you like to be part of that?" I can see all sorts of scams coming out of this too. It just gets even worse the more I think about it
Darin
00:39:42.037
So what do you think? Do you think there will ever be a truly zero human company? Head over to the Slack workspace, go over to the podcast channel, and leave your comments there