Win At Business And Life In An AI World

Exploring Trailblazing Business Models in the Age of Artificial Intelligence (Episode 166)

Vlad A. Ionescu is the Founder & CEO of Earthly; which was founded in 2020.

He is a repeat entrepreneur with an extensive background in cloud computing. Vlad co-founded ShiftLeft, the application security platform; created Lever OS, the first open-source serverless/lambda implementation; and co-authored the RabbitMQ Erlang client.

Vlad was previously a software engineer at Google (2x) and VMWare.

What you will learn

  • Discover what inspired Vlad to become a software engineer
  • Find out what cloud computing is and why it matters for businesses
  • Vlad unpacks the implications of AI for entrepreneurs
  • Learn more about Earthly and why Vlad created the business
  • Vlad discusses new and exciting AI business models
  • Vlad shares hard-won lessons he’s learned as a business owner in tech
  • Plus loads more!

Transcript

Jeff Bullas

00:00:04 – 00:01:39

Hi, everyone and welcome to The Jeff Bullas Show. Today I have with me, Vlad Ionescu. Now, Vlad is the Founder and CEO of Earthly. The company behind the popular open source developer tool Earthly, has a simple CI/CD framework. Now, really what we’re talking about is developing software that can be tested before it goes into the cloud and we’re gonna talk more about and remove the acronyms. But essentially, we’re talking about the technology that drives business today and can be trusted. He’s a repeat entrepreneur with an extensive background in cloud computing. We’re gonna talk a bit about that and what, why cloud computing is important to business. He obtained his bachelor degree in computing at Imperial College London, which apparently is one of the Top 8 Universities in the world. He originally got his degree, I think from one of the universities in Bucharest, so it’s in Romania. So we’re gonna have a chat about what inspired him to start. He’s a co-founder of ShiftLeft, created Lever OS and co-authored RabbitMQ Erlang client. And we’re gonna find out what that rabbit is about. He was previously a software engineer at Google twice. I don’t know why twice maybe they liked him and didn’t like him. I don’t know. We’ll talk more about that and he lives on the north side of the Golden Gate Bridge in San Francisco and looking forward to hearing about what inspired you to start to become a software engineer and start to get into becoming or starting a business around that.

So, welcome to the show, Vlad. It’s an absolute honor to have you here.

Vlad A. Ionescu

00:01:40 – 00:01:44

Jeff, it’s a pleasure to be here. Let’s get into it.

Jeff Bullas

00:01:44 – 00:01:53

Okay cool. So let me ask you the question, I always ask, what inspired you to become a software engineer?

Vlad A. Ionescu

00:01:54 – 00:03:51

Yeah, you know what, it was an accident to be honest, you know, it was interesting. So back in high school, I really liked physics, right. And I was in the national team in Romania. I was one of the reserves in the national team and just for context, the Romanian team is really strong, like we would get four gold medals and one silver every year or something like that. And as part of doing physics, I wanted to go to college and then the college for physics in Romania wasn’t very competitive and that sort of drew me off like I wanted something where I could be around smart people. And then I thought, you know, why don’t I go to this engineering school, the Polytechnic University of Bucharest and, the most competitive course there is the computer science course. So I applied there. I got in and that was sort of the beginning. But really, even before that I started, you know, playing around with code. I did that as part of my, just as a hobby. I didn’t realize I was any good at it. I thought people just like doing those things. But over time I discovered I had a special sort of knack for it and, I built some interesting pieces of technology even in my sort of high school years before university and sort of later on discovered. Oh. You know what, this is actually really special. Like people don’t actually do these things. But, yeah, then I went to Imperial College and learned just more and more about the interesting parts of software engineering. And it just kept me going. It’s this field where a lot of things evolved very fast and it’s really hard to get bored, with all the technology that’s going on right now. And, you know, it just kept me, kept my mind busy and entertained for a really long time and I just love doing it.

Jeff Bullas

00:03:52 – 00:04:08

Yeah, it’s certainly for me, I tried to, you know, use basic programming in the ‘80s and realized I was shit at it, really. So, I decided not to pursue that. I think you gotta be wired a certain way to be a software engineer, don’t you?

Vlad A. Ionescu

00:04:08 – 00:05:42

I think so. Yeah, I think it comes from my dad who is a perfectionist and somehow he built that into me in a way that is maybe both healthy and unhealthy. And I think as a software engineer, you have to be kind of obsessed about perfection and getting things right and being sort of standards driven and all that. And I kind of like the detailed work. That’s something not everyone really enjoys and maybe people find it boring and maybe I did it, I did find it boring initially. But as soon as, you know, I got to be able to do more things with computers, I got to program more complex systems and so on. Like the world opened up to me and it felt less and less sort of, boring in that sense, I mean, maybe the boring part was only for a couple of years in my high school. But, really, after that everything opened up and the possibilities were kind of endless. Like every time I was thinking about how I do like, I know this little bit of the field. Not a lot, but what can I do with it? You know, like what are the interesting sort of applications I can build with this little piece of knowledge that I have? And by building those things I then learn a bit, a little bit more than I could build even bigger things. And so on that sort of, spiraled into just more and more understanding and knowledge of the field that gave me more appreciation for it. And, yeah, I really love it, you know, today it’s been a really sort of a gift from nature, I guess.

Jeff Bullas

00:05:42 – 00:06:19

Yeah. And that’s, I think, that’s what’s authentic to you, important to you and also what you were basically born with, which is fantastic. So, the thing that interests me too is because software for me is so abstract, it’s just lines on a screen. Was there a moment where you actually started doing that and then produced something, in other words, you wrote code and then it produced a product or was there an aha moment when you did that for the first time I went, I could make, I gave it the code and then it produced what I wanted. Can you, do you remember that?

Vlad A. Ionescu

00:06:20 – 00:09:12

Oh, yeah, absolutely. So, for me, my journey started with just a bunch of very simple scripts that it was like an automated test for that just testing the knowledge basically, like providing a bunch of like multiple choice sort of problems and then you would have to sort of select the right answer and give you a score like the most basic thing. But that sort of, you know, it sort of inspired me about the possibilities and at first it was funny but it was more for me about impressing my colleagues that in high school, it’s like building these little demos that sort of they couldn’t build themselves. Like, we had this sort of computer science class even in high school, like programming class. And most people were just thinking about the basics, you know, how to do a for loop or like the basics of computer science and programming. And then I was just showing them interesting things you can do with graphics and like 3D stuff and all that. And initially, it was just about impressing my colleagues. I didn’t realize it was actually really meaningful for me to learn all these things and just to build the knowledge about what, you know, computer science is all about and computer engineering. And yeah, so like, it kind of grew from that simple, you know, from that simple initial journey of mine really the way I felt I started building something more meaningful is when I started working for a real company, you know, when it started, I started shipping things to production, there were no longer demos or things to impress my colleagues. They were like real meaningful things. And, you know, one of the first things I did was RabbitMQ, it’s this, I was in the early team over there. Is this piece of technology that allows servers to communicate with each other and it helps, you know, for example, banks to keep track of the stock market or it helps, you know, different servers to coordinate different activities like in a social network. There are many sort of vast applications to why servers need to talk to each other. But that was maybe one of my sort of early contributions. I worked on some of those pieces, some of the core components and I built some of the technology needed to achieve that. And that was maybe the first time I felt like what I was doing was meaningful. And also because it was open source, I could see how people depended on it. It was not just something, not just a prototype for a demo. So anymore, it was the real deal. I was making a difference.

Jeff Bullas

00:09:12 – 00:11:35

Yeah, so that area you talk about is cloud, especially cloud computing. And, so everyone’s heard of cloud computing, I might get you to give your description of what it really means. But what it sounded like to do with that RabbitMQ was that you allowed servers, computer servers in the cloud to actually talk to each other. So everything kept updated and synchronized. So I suppose everything is a point of truth, but even distribute in a distributed way. So cloud computing wasn’t trusted because it was in the cloud, no one could see it and everyone didn’t trust it. So we used to get our software delivered in boxes with CD roms and we used to load them up and we used to have our own, you know, head of IT, you know, head of IT in each corporation or government department running their own server network. Whereas cloud computing is where you actually let an organization like AWS or Google look after it for you. And now we’re at a point where we didn’t trust the cloud initially, very much, especially banks and everyone else because they didn’t have control. Whereas today, basically, almost everything is in the cloud. And I look, I, when at the end of the podcast here, I download it onto my server and I’m not relaxed until I’ve uploaded the recording to the cloud. And it’s to Google. In other words, I, because I know they’re gonna be better protection looking after my information than my little computer here that could have a little, you know, shimmy shake and breakdown this afternoon or this morning, which has happened before. But if I’ve got it all in the cloud, then it’s, I know that there’s systems processes, securities, software engineers, 24/7 in a dark room somewhere and looking after my data. So cloud computing has come a long way, hasn’t it? Maybe you can tell us a little bit about cloud computing and how important it is for business and technology because technology now drives business. Like, look at all the tools we use. We’re using technology that someone had to write code for. It’s called Zoom. So tell us a bit about your insights about cloud computing and how important it is, what it is. What’s cloud computing in Vlad’s terms and also why is it important?

Vlad A. Ionescu

00:11:36 – 00:14:52

Yeah. Well, I feel like this has evolved over time, you know, like cloud computing in the early days was something a bit more mysterious. There was no playbook as to how to manage it and how to deal with it. And so many organizations were maybe not so trustworthy of, you know, adapting it. And yeah, what is cloud computing? I think the easiest thing to imagine is, you know, someone else’s computer, right? It runs in the cloud, it runs, it’s managed by the big corporations today, AWS, Microsoft and Google are the main sort of providers of cloud computing and then you basically outsource your computation, your hosting your data, holding that to those providers to hold that for you and also serve your customers, you deploy apps to those cloud vendors and you are able to then serve your customers through them. The way this has evolved over time is that initially, many organizations did not trust each other with that data. And so everyone tried to build it in house. And in the early days, there was no good sort of playbook as to how to manage it, how to run it. And even today, it’s really complex. It hasn’t gotten easier, maybe it’s even gotten more complex with all the technologies out there. But we’ve gotten better at managing vendors and trusting AWS with our data, trusting that they have the right processes in place to protect our data and to provide the right level of assurance that it can support, you know, even the most secure applications, even the most demanding of applications. And even the government sort of applications, there are things like GovCloud and such like even the most sort of compliance minded organizations have adapted the cloud. And that is a big difference to what the world was like maybe 10 years ago or 15 years ago. And the risk I think is still there, but it’s just we’ve learned as an industry that it’s better to concentrate our efforts to give this kind of, sort of the stewardship of the cloud to the professionals like the AWS’s of the world, the Googles and Microsoft, they have the expertise in house to manage the cloud in ways that no smaller company would be able to do. And so the barrier to entry has been reduced dramatically like everyone is now playing to the cloud. You can be the smallest startup ever and it’s very easy just with, you know, a few 100 bucks, you can just get going and that’s a big shift forward for things like reliability and security that were not maybe accessible some time ago. So, yeah, the future of the cloud is really bright and a lot of innovation has been based on it. So, it’s really an interesting space.

Jeff Bullas

00:14:53 – 00:15:45

Yeah, it is very interesting. I’ve been involved in tech since the mid-80s when PCs were out and they weren’t even connected to the internet. Okay. So you had your own network, first they were standalone. Then we connected them in the house. IBM invented the first one of the first networks called Token Ring. And then there’s a more open source Ethernet emerged. And then I remember having discussions with even like, I think it was Amazon web services early on. It might have been in the early 19, late 1990s and you had to go and basically sign big contracts where you had to buy your own servers. It costs you $20-30,000 to do that. Whereas now in the cloud computer services are like a monthly subscription which you can start as low as you like and then build it up over time and scale it as you need to.

Vlad A. Ionescu

00:15:45 – 00:16:10

Exactly. That flexibility is magical by the way, like there’s no way, you know, companies could grow so fast these days from like a really small startup, like for example, what Clubhouse was doing over the pandemic, right? Went from like an unknown company to this huge space that everyone was into. I don’t think it’s easy to do that without the cloud.

Jeff Bullas

00:16:10 – 00:17:05

Yeah. So we wouldn’t really be seeing the explosion of Silicon Valley entrepreneurs unless we have that cloud. Now, this raises another interesting thing too, I think it is important to raise. We’re in the middle of the biggest generational change in tech that I’ve ever seen and I’ve been around in tech for approaching, shit, 40 years. So I saw the PC revolution happen and I was in the industry when it was happening, which was most exciting. We all got excited by spreadsheets and, you know, word processes. And whereas today we have a new tech emerging artificial intelligence, AI. Now, the interesting thing is, AI would not be happening unless we had cloud computing capability.

Vlad A. Ionescu

00:17:06 – 00:17:17

That’s right. Exactly. And I would say, it would not happen without the internet either because all of it has been on the internet. So, that’s an interesting sort of prerequisite.

Jeff Bullas

00:17:17 – 00:18:31

Yeah, exactly. So, ChatGPT runs on foot. Well, I heard back in, ChatGPT has only been given a human face to democratize AI for the general population which happened on November 30 last year. And yet we feel like we’ve been living together forever and it’s only been about four or five months. So, and it’s fascinating that the reality is that behind AI, it got trained on the internet on big data sets. So in other words, generative AI, in other words, it learns by generating information, learning and continuing to learn. I maybe haven’t said that exactly right. But I think I’m close to it anyway. So, Microsoft powers ChatGPT. And I think the first iteration of it back in November was about nearly 3000 computer servers that were powering ChatGPT and I’m sure it’s a bit bigger today. So let’s have a chat about the importance of AI and software engineering now and where we’re up to that because AI is now starting to write code. So let’s have a little chat about AI and software and where you see that going and the implications for entrepreneurs.

Vlad A. Ionescu

00:18:32 – 00:22:17

Yeah, I think there’s gonna be a big shift in the way we write software in the next decade or so. It’s not clear when exactly the AI will be able to just fully take over the engineering responsibility. It could be in the next two years. It could be in the next 20 years right now as it stands today. It can do this in one step. You know, you ask a question, you get an answer but it cannot yet write full applications. Maybe you can write a small demo, you know, show how something very small can be programmed very in a very small manner. But once you go to like bigger serious apps that would sort of maybe power businesses, it’s not ready for that yet. You could maybe run it in a loop so that there are projects out there that actually run ChatGPT in a loop, basically mimicking the way a human works like trial and error. You know, if you try one thing, you sort of don’t get the right result. You try again and try that over and over again until you maybe you get the right result. But because there is still some inefficiencies like inaccuracies, I should say, as part of that process, sometimes that initial small error can amplify as part of a, you know, as a repeated pattern. And so we’re not quite yet at the point where AI can fully write like a fully fledged application that could power a business, but it is certainly a tool that changes everything at the very least, you know, like developers plus AI is a significantly better combination today. And that will only continue to improve. So just since the launch of, you know, ChatGPT, if we’ve gone through some iterations of the model that there have been significant improvements to it. And already we can see some just across a few months of work, we can see significant differences in the quality of the output. Now, if we scale that up to say 5 to 10 years, we can sort of draw the line and the trend is very interesting. It’s possible, [inaudible] extreme that software engineering is going away as we know it forever and we just have basically product people being the engineers because they no longer need the sort of nuts and bolts knowledge about how the software necessarily works. They can just use the AI for that. So that’s maybe one extreme. And the other extreme is where I think at the very least it is a very powerful tool that makes developers significantly more productive than they ever could be before. And it’s hard to say where exactly on the spectrum we’re gonna land. But I think, so far we haven’t, we maybe can address the problem of how to write a simple application with AI. We haven’t yet fixed the, address the problem of how do you run that application once you have it, how do you maintain it over time? And sort of watch it work well in production, watch it sort of sometimes fail and then you have to go back and fix it, address the issues, address the scalability of it and all that comes with, you know, managing an application in production, which is the other sort of 50% of, you know, being able to power an app through AI and that part, we haven’t done, any progress on just yet. It may be possible to do in the future but it’s still kind of unknown at this point.

Jeff Bullas

00:22:18 – 00:22:34

Yeah. So it’s really fascinating. It’s great to watch it and how it does software. And yes, like you’re saying, it’s not providing a school full solution, but it’s amplifying programs, isn’t it in terms of writing code faster.

Vlad A. Ionescu

00:22:35 – 00:22:36

Exactly.

Jeff Bullas

00:22:37 – 00:22:43

And is it also able to test faster as well? Is that part of the equation?

Vlad A. Ionescu

00:22:44 – 00:23:54

It can help you write tests faster but not necessarily run them faster and the way it works. To anyone who hasn’t sort of played with this kind of technology, there’s basically, it’s an automatic suggestion engine where as part as you type code in your sort of editor on your computer, like this extra text appears a suggestion to augment your program and add sort of the next thing you are going to type is suggested by the AI and by doing that alone, you can sort of use it to be able to generate like much more comprehensive testing that maybe you would have the patience to do. Instead of you maybe give it an example of one test that you would write and then the AI comes up with another 20 like that, but just testing different edge cases. And so it is incredibly useful in these various areas. It doesn’t necessarily operate completely independently just yet, but just the fact that it makes individual developers be that much more productive is just fantastic.

Jeff Bullas

00:23:55 – 00:24:45

So there’s sort of like two camps to AI, one which says, you know, it’s gonna take over the world. So watch out that we humans will just become a subspecies of AI robots. And then they go, well, it’s just gonna amplify us as humanity. I’m more in that camp than the other camp. And then we get experts out there so called, like Elon Musk who signed this document saying we need to pause things for six months while we write rules for it. In the meantime, Elon Musk starts a generative ChatGPT clone business. Elon Musk is just talking shit, really? He’s saying pause it while I can build my software and catch up.

Vlad A. Ionescu

00:24:45 – 00:25:29

Yeah. Exactly. It’s kind of strange but, I don’t believe in this kind of, you know, hold, you know, hold your horses. Don’t build anything else just yet because it’s dangerous, right? I don’t think we’re that close to that kind of scenario. Maybe it’s possible. I don’t know, in the next 50 years, maybe AI will take over the world, it’s a real danger. But we’re not nowhere near close and I think, just, yeah, that doesn’t feel fair for someone to be able to sort of influence this kind of a widespread decision in the industry while at the same time playing catch up. So, like, it’s definitely a conflict of interest there.

Jeff Bullas

00:25:29 – 00:25:38

Yeah. Well, I’m just gonna say something to Elon, I see you Elon. I know what you’re doing. Okay. So, watch it be very, very careful because we know what you’re up to.

Vlad A. Ionescu

00:25:39 – 00:25:41

That’s right. Yeah. Exactly.

Jeff Bullas

00:25:41 – 00:25:58

Cool. So, let’s talk a little about Earthly. Let’s, so tell us what Earthly is in words and sentences that I will understand and our audience would and why you created Earthly.

Vlad A. Ionescu

00:25:59 – 00:27:31

Yeah. So Earthly is, by the way, unrelated to A I. I know we’re sort of switching topics here but it is a tool that helps developers be more productive. And the way it does that is by allowing you to run a developer specific automation significantly faster and more efficiently. And that developer specific automation today is called CI/CD. It stands for Continuous Integration Continuous Deployment. It’s the stuff that automates the journey of a piece of code from the moment you type it in your, on your computer to the moment it lands in production and that process runs many times in a single day. It tests that your changes are correct. It tests that everything works as it should before shipping to production, it also packages and compiles the code and does all that magic stuff that puts together the application that runs on the servers. Our technology is able to dramatically improve the way that piece of technology works and that results in 2 to 20 X improvement in the performance of that system. And that translates directly into developer productivity gained back. So you can achieve more with less people or you know, be able to be much more productive than you were before. At the end of the day, that just means less bugs and more features with the same headcount in terms of like engineering organization. And that is the part of the technology that we deliver.

Jeff Bullas

00:27:32 – 00:28:00

So what you’re, what I’m hearing is that the process and framework that you use with Earthly basically amplifies and accelerates humans as developers of software. And on top of that, then it actually makes it more accurate as well. In other words, it solves problems quicker and corrects them or be able to test things. But so it’s got that attention to detail designed in. Is that correct?

Vlad A. Ionescu

00:28:00 – 00:28:47

That’s right. You know, these things historically have only run in the cloud, this kind of automation. And part of what we do is we allow it to also run on your computer. And that feels like a very sort of fine difference. But in actual terms that actually turns out to be very important because you cannot take that automation and play with it on your computer in isolation. And there’s a level of guarantee that it works the same no matter where you run it, like whether it’s your computer or your colleague’s computer or the cloud itself, and that has tremendous advantages in the way these things are developed and the developers can then be much more productive in the way they work with these things. And, that is sort of the novelty of our solution.

Jeff Bullas

00:28:47 – 00:30:29

Right. Yeah, it’s, the reality is that we are using software so much in business now. In fact, almost everything. So this call is we’re recording it on software which is driven by algorithms. And I’ve, I now almost implicitly trust it more than my own computer and myself, but it’s really interesting on the AI side of things, for me that AIs has got like this broad spectrum of AI which is sort of general AI, generative AI which works for everyone. But then we’re ending up with soft specialties. Well, we had verticals and specialties for software, but now we’re ending up with AI that actually looks after AI that can help verticals. For example, I had a chat last night with the founder of Assembly who it’s just using AI to actually capture meetings, distill meetings, summarize meetings, create agendas, write the email for it. So, and then, businesses run on meetings a lot of the time. So who wants to take notes? And then we humans are very imperfect at taking notes. So, AI is certainly, you know, you’ve got the big cloud, you’ve got the big, you know, computer companies and you’ve got this big three which are AWS, Amazon Web Services. You’ve got Microsoft which powers ChatGPT and many other companies and then you’ve got Google itself.

Vlad A. Ionescu

00:30:29 – 00:30:30

Exactly.

Jeff Bullas

00:30:30 – 00:31:18

Yeah. So, in terms of open source, we haven’t talked about that. But, and that has some implications for AI as well. In other words, people write proprietary code and then companies try to hang on to it. So no-one knows what the code is. It’s hidden from view. Whereas there’s this debate going on even between open source AI and proprietary AI. Would you like to talk about that a bit and what your views are on that? Because we ended up with software that’s open source. Linux, which is a, you know, and I think most of the world’s servers I think run on Linux now these days, I think as well.

Vlad A. Ionescu

00:31:18 – 00:34:29

The vast majority of the servers I people don’t realize, but Linux is actually the most popular operating system in the world. Like you don’t see it every day in front of you, but it’s actually powering all the services that you use behind the scenes, like, you know, Google and, you know, just about everything that you do online is powered by Linux and it’s open source. Yeah. So, yeah, it’s a good question, you know, historically, open source software has been more sort of impactful in sort of a, in a widespread manner. So there’s maybe this decision to make whether as you are the author of an AI tool, do you want your AI tool to become sort of a standard in the industry on top of which you build other sort of commercial applications? And then you would have widespread impact with your open source contributions. But maybe there you have to think about how you want to monetize it if you wanted to sustain it. Not every open source project works through the donations. Like I think Linux Foundations and things like that. They have been tremendously successful, but the vast majority of open source software needs to be powered by something, some commercial sort of effort to be really sustainable. And, then on the other hand, you have closed source software where it is primarily designed for it to generate money and that’s why it’s closed source. You want to hide your moat, you don’t want to give away the secret sauce. Exactly. And, so there is a debate because once you do close source, you get less developer love. So it depends on what exactly you target with your business. Sometimes it’s worth targeting the developers and building the infrastructure of AI, you know, the stuff that powers everything else. And there are many sort of playbooks by which you can actually turn that into a monetizable product as well to sustain it. But also you could be building something very sort of end user driven. So you don’t, you’re not appealing to developers, you’re appealing to like, you know, the consumers and so on. And I think in those cases, it makes more sense to be closed source, right? Like at that point, there’s no benefit to you necessarily to open source your technology to give away some of that IP there’s more benefit in you sort of holding tight to that and sort of using that for monetization and then anyone who wants to compete with you has to recreate that technology from scratch basically. And I think it all depends on exactly what you’re going after as a business. It might vary depending on whether you’re building the infrastructure of AI or you’re building the end user sort of consumer oriented type of business.

Jeff Bullas

00:34:30 – 00:35:50

Yeah. And this raises the question of, well, big question, which I’ve been thinking about for a while is that basically the internet opened up new digital business models. And we ended up with, you know, a monthly subscription instead of, you know, buy your software, buy the box to renew it every year or two. And of course, we all end up with really outdated computer systems because no-one wanted to spend the money that Microsoft wanted to charge you for updating 300 computers. Today it’s very different because everything is updated in the cloud. So we’ve had the evolution of business models driven by technology. The thing that intrigues me, I’d be interested if you have any thoughts on this, is what are some of the new and emerging AI business models? In other words, technology driven by AI or AI driven technology that is gonna change the way we do business because you must be watching quite a few companies that are developing software. Do you have any thoughts on some of the, where you see business evolving into more AI business models? I’ll be intrigued if that crosses your mind.

Vlad A. Ionescu

00:35:51 – 00:38:11

Yeah. So far, what I’ve seen is that the companies emerging right now are just using the business models of the internet. So, you know, the subscriptions, maybe the pay per API call so that the call to the programmer interfaces of those services that give you AI and I think, we haven’t really figured out exactly what maybe an AI driven business model can be, but it could be either around the value that you get. So, for example, you know, if this AI can make me, I don’t know, pizza, right? You would pay for the pizza. That’s the result of that sort of interaction, right? Whereas in more infrastructure terms, if you were to build something that is meant for developers, a pay per use is more typical in that environment. So instead of paying for the end result of pizza, you would pay for, you know, the ingredients that went into it, you know, the flour, the dough and, you know, the toppings and so on. So I think and then there’s the subscription model where for example, you want a continuous service to something and so for example, maybe this continuous service is, I don’t know, generating ads every month and then running them for you. That makes sense because it’s a very continuous process. You just, you don’t just get the value in one day or you don’t necessarily get the value just by interacting with it like on and on. It’s something that just works automatically for you. And, so subscriptions make a lot of sense in that case. So I think a lot of the models that power AI will be simply at least so far very driven by what we’ve learned from the internet era, the software as a service platform as a service and so on. Although, you know, time will tell exactly how things evolve and how these agents become much more independent and, you know, what will come out of it as a result of the sort of business aspect of it and the pricing of it.

Jeff Bullas

00:38:12 – 00:39:23

Yeah, it’s gonna be fascinating because we are trapped in the old models because we, as humans get trapped in templates that we’re used to and change is hard. Change is scary. So, because this intrigues me in terms of what does an AI business model look like. Well, it could be in thinking about this a little bit, I believe it’s gonna be the intersection of various AI driven technologies that will be combined into one product or a process. Let’s think of media, for example, AI ChatGPT can create content, AI can create a video or an image, AI can detect what works and what doesn’t when you release that ad to Facebook or Google, then there’s gonna be creating of the landing page for that ad to actually go to once people click, then it could be the delivery of it. So I, so I’m thinking about it a little bit. I’m sort of seeing, I’m thinking that maybe it’s gonna be the synergy of multiple intersecting technologies driven by AI that create a business model that could be almost automatic.

Vlad A. Ionescu

00:39:24 – 00:41:17

Yeah. It’s possible and you know what something that’s gonna shift a lot? It’s gonna be the cost of content production will go almost to zero, right? And in a way, we have some experience with that model because in a way social networks are like that, you know, you give away the social network for free and people just create content for you for free all the time, right? And at a very large scale, so we have some of that experience. Now the quality of that content will be of a different nature when we combine it with AI. And you can think of not just social networks, you can think of like, you know, personalized movies that the AI generates for you specifically to match your preferences and the type of drama that you like specifically as a person. Or I don’t know, songs that you know, sound like your favorite artist, but maybe your favorite artists are no longer around or something or, you know, all these interesting effects and I think really the net result of this will be that the bar for quality of content will go up. So the because the the price of generating content goes down, then you can have mediocre content all you want, you know, that’s gonna be sort of standard, but now the really high quality content will be the sort of the differentiator and all the sort of Netflix of the world and the sort of content generation of platforms out there will probably want to raise their bar, use these tools underneath to raise the bar, then sort of deliver on, just content we’ve never even thought about. It’s really hard to imagine, probably at this point. But, yeah, that, I think that’s gonna be like a very sort of dramatic shift that we kind of don’t expect as a civilization just yet.

Jeff Bullas

00:41:18 – 00:41:46

Yeah. And, what’s really fascinating for me watching, not only does AI help us amplify our humanity and create faster. On top of that, then I think what I love about AI when you put something and use the right prompts. And that’s the other thing that’s really important with AI is to give it the right questions and usually the best questions come from those that are experts already in their field because I know what questions to ask.

Vlad A. Ionescu

00:41:47 – 00:43:16

Yeah, you know, it’s possible that this will be a shift in, you know, just reducing the barrier of entry kind of like cloud computing. I went from this era when everybody was doing on-prem cloud computing, they had their own servers, their own racks, they would manage that on their own. It was very expensive to do. And now to the era when anyone can do it, like any startup can do it. And similarly with AI, you historically needed like, you know, very expensive experts like PhD’s in computer like in data science and AI specifically machine learning and all that. And now with prompt engineering becoming the de facto API of AI that barrier of entry is reduced dramatically so that regular engineers can do it, with which they can experiment with it. They don’t need to have a, like a Ph. D. in data science or anything like that. They can just run with it on their own. And the effect of that will be that just about every company out there will have an AI play as part of it. It’s already happening, right? We’re seeing every single week, just hundreds of new AI based applications being released and it’s just mind boggling just how much innovation there can be already despite being so early in the process of discovering AI.

Jeff Bullas

00:43:17 – 00:44:25

Yeah. And everyone gets access to an AI platform that is basically enterprise. NVIDIA have just announced an enterprise level or enterprise access to all their servers with their access to an AI platform without actually having to build it yourself. And ChatGPT sits on that, but this is an enterprise layer. So we, like you said, we’re already starting to see new business models emerge that we are reducing. Like you said, the barrier to entry for people to start their own business in a niche that could be a tiny specialty. And the implications for search are also huge because we can create content that ranks for the keywords that we discover that have got keyword violence worth chasing and we can create more content. Then the other part is to optimize that content for conversion rates. So my team is leaning very much into using AI for content creation but driven by a human editor, which makes human and voice. And it’s mind boggling so we can create 20 pieces of content in an afternoon.

Vlad A. Ionescu

00:44:26 – 00:44:28

Exactly. Yeah.

Jeff Bullas

00:44:28 – 00:45:06

And we’ve done the research about what are the top key phrases we want to be looking for that have got enough volume to actually be profitable because, and we’re looking at 1000 plus global and I’ve been playing. Yeah. So we get, you know, my humble, the site gets about two million to three million search hits from Google a year. And we’ve got a range of verticals we rank for on the business software, business tools, business marketing, we’ve got AI marketing and so on. So it is fascinating and we’re actually experimenting to see what we can do.

Vlad A. Ionescu

00:45:06 – 00:47:29

Exactly. You know what we’re in a very similar stage with our content marketing efforts. Actually, here at Earthly, we had 1.3 million views last year and we’re projecting 2 to 3 million this year. And we’ve been doing similar experimentation with AI and in fact, as soon as some of, you know, ChatGPT was released. We already saw AI being used by our contributors. Like we have, we run this little program where we get external contributors to write for our blog and we saw more plagiarism as a result, I don’t know if they used AI for that or not, but it just, we just saw things being sort of copy pasted from other articles. And yeah, so like, so basically these contributors were paying them money to write for us. And, basically, they increase our, they write content that is driving, you know, our target audience to our site basically. And we have a certain standard for quality. And as part of that quality check, we also check for plagiarism. Like we Google the terms in the content and figure out whether they’re actually, you know, they’ve been plagiarized in any way and we, it was kind of sad to see in a way that maybe AI is contributing to this. But at some point, we also realized, you know, what if this is high quality content and we ourselves at some point cannot tell anymore whether it’s plagiarized or not. Does it really matter that AI was used to generate it? You know, if it’s useful to you. And, so right now we’re in this journey where we even use it a lot for iterating on our content and just finding better wording, making it simpler, improving the content overall. I think it’s, we can’t really trust it completely to write complete articles for us. Sometimes it goes off rails and, you know, the quality can suffer in some cases. But it’s very similar to software engineering. Like it’s this amazing tool that makes you more productive with a human in the loop that we never had before. And now we’re suddenly realizing we can no longer live without it. Like it’s this far hole.

Jeff Bullas

00:47:30 – 00:49:07

Yeah. And we’re taking a lot more control over our content by creating content that is built upon strategic selection of keywords and phrases that we want to rank for over time. Whereas before we were getting guest authors and then we’re trying to be jam that content into a key word or phrase that we don’t necessarily want. We don’t think it’s important, doesn’t have enough volume. It’s not the right target niche that we’re after. On top of that, then we use content to generate traffic to affiliate partners that we have as well. So it’s, we’re experimenting and playing with the creation of content. But then it’s almost like create volume first still quality, but then optimize it for conversion rate later, then optimize it for visuals on that later and guess what Google loves new content. So, republishing every couple of months is actually a very good SEO practice. So, ChatGPT and AI can help you continue to iterate articles and make them better and better and better. So it’s almost like, so we’re starting with core content and now we’re gonna just evolve it over time and also it’ll send a signal to Google. Google tells us that they can detect AI written stuff. I’m not sure they can yet. So, yeah, so how much can they detect? Because if you’ve got enough humans that write the intro, edit a bit, it’s not actually AI generated, it’s written by human and AI.

Vlad A. Ionescu

00:49:08 – 00:49:10

Exactly.

Jeff Bullas

00:49:10 – 00:49:53

So this is where the business models start to where you get the different integrations and intersections of multiple technologies all driven by AI now because it’s gonna be woven into everything. When I watched the, you know, Assembly’s, an example of them using AI to basically record meetings, distill meetings, write an email from the meeting, summarize the meeting. Now, they even can turn up as an avatar at the meeting and ask the right questions is evolving as well. So I think it’s gonna be digital business models that are not just internet driven now but AI as well as, it’s gonna be fascinating.

Vlad A. Ionescu

00:49:54 – 00:50:32

Exactly. It’s gonna be a prerequisite it’s kind of like in the mobile era, maybe you were working on the most interesting app on the internet. But if you didn’t have a mobile app, like I, if you just had a web experience, it wasn’t, it wouldn’t take off. I feel like AI will be similar, like everything that we do as part of, sort of innovating on top of, you know, innovating in general, in the space of cloud, and mobile and so on, they will have to have an AI component to make them up to date with what is going on right now, like you’re gonna be significantly behind if you don’t do that.

Jeff Bullas

00:50:33 – 00:50:58

Yeah, exactly. So, just to wrap it up, what have you learned by being an entrepreneur and starting Earthly as, and you know, what are some of the big lessons as a technology business owner? And that lives and breathes software which drives the internet and drives AI as well? What have you learned along the way? Just sum up what you can share with our listeners and viewers.

Vlad A. Ionescu

00:50:59 – 00:52:25

Yeah. I think for me personally, it goes back to the basics, like work on something that really, really drives you that you really, really like. And Berkeley is my second successful company, but maybe my, let’s say fifth attempt at startups and I realized in the process of like e-trading, failing and then try again that you have to go back to something that you really, really love otherwise, it’s just, it’s maybe too easy to get distracted by something that feels interesting for a day or maybe a month or maybe even a year. But if you’re gonna do this for like 5 to 10 years, like for example, an Avenger backed startup, your heart has to be entirely in it. And, so for me, for Earthly, the main prerequisite for me starting this company was that I would be a target user of this technology. And then I would be in the feedback loop. I would know exactly what is not working because I’m also a user, I’m also the product person, right? I can sort of iterate much more quickly as to what the shape and form of that technology should be. And so, and my main sort of takeaway is as my, as I learned about how to build the company is just do something you really, really love and are passionate about. And I think that has worked historically really well for me.

Jeff Bullas

00:52:25 – 00:52:47

Yeah. And I would totally agree because if you are doing something, number one, you’re good at number two, you love doing and solves the business problem, then you’re gonna be able to stay the journey, which because being an entrepreneur is sometimes a hard slog and sometimes, you know, an overnight success is 10 years.

Vlad A. Ionescu

00:52:47 – 00:53:16

Exactly. So, every successful startup has those stories about the worst days of their existence. They all have, you know, tough times if you’re not 100% into it, there’s nothing to fall back on to, like, you know, everything doesn’t feel, it feels like it’s not going well at that point in time. And, you have to just keep going. You know, if you, there’s the saying, if you’re going through hell, keep going and if you don’t have the passion to back that up, you just want to keep going.

Jeff Bullas

00:53:16 – 00:54:08

Exactly. And I think that’s I think it’s important for every entrepreneur to know that don’t just chase the money, money results from you being passionate about something and also having expertise and also taking the time to play the long game. And I think that’s very important. So I think you’ve summed up beautifully that is a core characteristic of being a successful entrepreneur, inspirational, wrapped in desperation and longevity. So thank you, Vlad, very much for sharing your insights. And we’ll try and our attempts to make software interesting for the greater population of entrepreneurs. So, but it drives the internet, it drives business models.

So it’s great to have a chat about it and have a chat about AIs along the way. Thank you very much. It’s been an absolute pleasure.

Vlad A. Ionescu

00:54:09 – 00:54:12

Awesome Jeff. This was a lot of fun. Thank you.

Latest Shows