T3 EP10 - AI Adoption in Highly Regulated Markets | La Hora del Tech

(upbeat music) (upbeat music) - All right, so I'd say the two guests, we have, there's a reason why we picked them to be. In our last episode, so it's not because he's the CEO and the boss, it's not because of that. Not just that, so Mike has 35 years of experience in background in healthcare, life sciences, technology and data, management. So he's been through many different situations in BL. He's an expert in everything that has to do with compliance with cloud. So they can actually give you, give us, give us a real perspective on what's really happening in the industry. There is a lot of it. We're going to answer all those questions. We're gonna get very deep into AI, real usage, how we navigate this regulated market that we work with, and then we'll be able to unpack topics that we hear very often. - For all the audience, Bill Santos does also Latin American Roots. So out as communityLT, give us a localized example of what proprietary data means for someone in Colombia. - So yeah, so, so I'll like say, we'll discuss something that everybody knows 'cause everyone has to do it. Everyone has to pay taxes. You have tax laws, I don't know, there's tax exemptions, there's all these other forms. So actually ingesting all of that tax form data, keeping current and making something that's an easy access tool for someone is a way to say, like, oh, like, we're pulling in all of this government data, but we're also finding these external programs or these offerings, private companies or private entities to enrich our data, so we're bringing these desperate data together to create something new. - As on my thing, a real world example here in Colombia is like, yeah, that character that is people, there's a company that gathers financial data and then give you like, like, credit scores. So they, you can actually go to, you know, and they're like, please, and say, hey, I wanna buy this. I wanna, I don't have money right now, but I can, you know, use my credit score, you some thing, you don't need a credit card, but you just need to be in that database. - Yeah, that's another example of like, they're the system of record, right? So that's a trusted score, everybody accepts it, and they make decisions based on it. So it's not that what they're doing is that hard, it's that they build all those, always sort of trust connections, you know, so that they'll stick around. Okay, so I want to change a little bit of focus here, because we want to talk about protecting or being compliant with data, which is a shout out that I don't think other companies pay attention to move, but we ask because we work in healthcare industry, we have been working with very, they very sensitive data, right? So we need to protect and not just protect, but also we are being very highly regulated, right? So we need to, like how to AI that are we going to be able to handle that regulation? It is very common because AI is everything, or they, unless you make a regulation for it, so how it's being handled right now. So if you look at AI and healthcare data, when I look at a bank contractor someone's doing you eat healthcare data, they carve out AI use rates, and basically say you're not feeding us to a model. If you want to feed it to a model, there's a different guy he's going to want to feed to your business to do it, because they view that as like, that's where this is going to make them potentially this enormous amount of money, and actually what more afraid of is that they don't understand how to make money look at it, so they're going to make what else gets their first. I don't want to say, I'll owe I pay $2 million something to data and someone will fill the healthcare version of them profit with it. So car contracts for proprietary healthcare data are carving that out and protecting it. History, however, gets for sale. And you see, you see a lot of things happening on the imaging side, where basically right now you have a workflow that started with, hey, I go to the hospital, I get a scan done, someone at the hospital, the radiology department looks at it and it hasn't hit yet. And that shifted probably about 15 years ago in the US to do a scan and then a doctor in bringing you deal looks at it and tells you what do you think C-Sees. And then that's turning you to Carthlux at it, this is what it is. So you're seeing this workflow that used to be, I need a guy on site because I didn't have a way to send the scan to, okay, I have a way to send the scan, I'm going to send it to a cheaper place to, and now I'm going to give it to an AI and be done. And the reason we're able to do that is history gives you the ultimate training set. And they know with every scan, like I've done in the last 15 years, they've got it, I mean, know what happened to that person. And you can go train the AI on it, probably once you've done it, you're good. Yeah, so that history turns out to be valuable, but in just interesting problems through that guy selling that data, it's valuable once. Doesn't have a scan to use. - So actually on that, we're at Udus and actually for me, I work in at source for around seven years. You get used to talk about healthcare, like science's sensible private data, but for our Colombian audience, and since you're an expert, you've been working with healthcare life sciences in the US ecosystem for many years, but I also, you've been in Colombia for more than 10 years and you're a great ambassador. So could you explain or do a comparison between the US healthcare system versus Colombian healthcare system? - So I'll talk about a data compliance thing first, and then try to explain how ridiculous US healthcare system is. So it, plus healthcare has a rule called HIPAA, which is, it was supposed to be about this problem you had, like when you try to switch doctors, it would give you your records, and it was basically a way for a patient to say, "Give me all my records." What it turned into was a way for these guys to be even worse about sharing this stuff. And one of the things that's codified is if your health information gets out in the open, the fines are unbelievable. And the fines are so high, and it's per record, it gets exposed, that basically hospitals look at like a HIPAA violation as like back to just close the doors. So they get very careful about all that, who a little share data with, and it's created this whole industry of the identification. I'll contrast that with, I was briefly talking to a minister of health here in Colombia eight years ago, and I think gave me the entire department's healthcare records on the drive. So slightly different focus on patient privacy. I gave it back. - You've paid any kids? - I didn't look up anything I do. I did. I looked at it like someone who's dealing with HIPAA, and I'm like, "Why did you give me this toxic thing?" But yeah, it's different regulatory. So okay, now to explain US healthcare, we wind up having insurance, usually paid for by the employer. And we created a particular tax reason by that. It's a good thing. You know, it started into the world where you're insurance comes or you get everything you're trying to do. And what their rules, try to keep the rates on the parole, you know, sort of get it in your way every once in a while. And then we also had a call at the AMA, there's a association of all of the medical doctors. And they create reasons for some points that you don't wanna have. So if I get sick during Columbia, you know, the flu, I go down to the pharmacy or I run it back to eat, you know, my drugs, go to my apartment, and I'm here. I can't give you that in space. Like they want to get a little employment. Several of these hang out in the waiting room, take time off, so that they can write paper that I take the pharmacy and get it filled. I have the insurance company go, I don't wanna pay for this. (laughing) And pharmaceuticals are trying to get this in the US and like the way we access it is beyond ridiculous. And it's resisted all of that pretty bit of technology. And, you know, my own experience that is just, you know, doctors trying to score an extra appointment. But it's truly, It's really broken and have easier access for that stuff here and and actually you know This works better here Better to work and I know it's I have an easier time accessing healthcare here Now we might have some great in like you know in New York or Philadelphia That's like you know a reverse or top in plots creator that something that nobody understands that you know There's too old or all through there, but like for normal healthcare stuff. It's easier here. I I agree But okay, so now that you explained to us how all that Complexity over health curse is this and because we in the company what we Helicator data the United States. I would want to ask and Again, maybe a question how difficult is it to comply or to implement Infrastructure or a system that help to comply with all of those Complexity controls that you need to follow to be able to work properly with the data So it's actually not hard, but it's very to make a mistake and then let data out We really need to focus on Well, what are the what are the laws and you build your safeguards there and then just finish your day cards there You have build monitors in case someone hurts something else opposed to do You know platforms and security and think of clients Was or always the logs looking for our client Almost to still be engineered from so such that engineer if we write some tries to send Sent data outside of the network there's alerts and there's blocks that stop it and immediately think appliance team knows Hey our application just right to do this and so it it really is around your system design and your security design That secures it Having worked both in education tech where you have to protect students and filburns names and The grades of their parque dates and all of that as well as working in the healthcare industry It really comes down to Want only to limit the access not only of the developer, but also of you know the Access of the API that's being written it's like it shouldn't be able to read the whole database You can even go down to like grow specificity if you want to go all the way to the per this Table limitations what it can read what it can to beat what it can Edit and then network controls that limit access geo blocking so you know Countries that are amongst the steel data from Especially like us us whole carry it because it has all of your Identifications to fuel your identity your money. So it really comes out to like designing this system and building monitors on top of that So to make sure if something is attempting to do something wrong you catch it Um such that you don't actually get to find the data having been stolen from you So Miguel on that and then you describe Unless how you know the developer and human beings work around this very protect data, but now Um with the viable and then with it unpredictable nature of LMS How do you deal with that in a daily basis for challenges? Have you had in recently using AI? Uh, I mean the biggest challenges a eyes are very smart and so um Uh really that they could catch on to things and they can find better to they like oh I need this one extra piece of data to do what I'm doing, but I can't get there But then there's smart but in a dumb way Where they'll be like oh, but I found this key store that has passwords that it'll actually give me the access I need So I'm now going to just grab that password and go do something that I'm not supposed to do So I was supposed to only use this one access token And it's like that that's really the challenge is like putting the safeguards on what you actually make available Uh to the AI uh agent that you're using to develop Exactly and then that brings me to you uh I recently had this with the fable 5 thing um that now is repeating and everybody's like Steve talking about it's not stout a bit of a little bit uh for the I don't know 24 hours we get access to So I tested it and then I was like it was he was a POC all right so it's like what's it wasn't a real party was a POC and then um It's fine. Yeah, yeah, you're dead as fun as well. So So what I let it like pretty much I asked it hey I wanted to find all the security Um like issues uh because that that's what it was created for initially right like mehthas 5 or 5 and then what it did is All right, so I need to pretty much scan everything and I didn't I pretty much moved all my personal keys but I found a way to connect to the container running in AWS and then the container was had access to the database and I didn't I didn't know it I guess but I mean I I didn't know it but I didn't know that the that model was able to connect to the container and be able to create a database on my behalf So it did a great job but think about if it something goes wrong, you know the list that they have made or move it in a place that we should and kids um actually this all this was that you just described or there's so uh let's say uh The strict that uh you can go to jail if you download the wrong dataset and your local machine And it that the difference is like we used to build systems that We're trying to keep you from accidentally doing something wrong Right though that they were they were card rails on on humans to like not do anything stupid Yeah, so if someone's gonna go get some PII and try to pull it out of the company like They got to get to do them for work for that like it's and and most of you know most of the like that you know someone next no one's accidentally gonna go find your Doctor image and then scratch around for keys in order to get into something they weren't allowed to get into In order to grab a piece of data like a human that's not you know an adversary is not going to do that But cloud will Like all the time like I I had I was doing work with cloud and I gave it this service account I could only read things because I didn't want it to get excited You know delete the entire cloud that I was putting on and like One of the one of the machines that have it on Like I also use it to lock it for my Gmail to go pull attachments down and like it couldn't do something You know, oh wait, but I found your credentials over here and I'm like you know kicking the hammer out It's like stop it stop it stop it for me. No So if that happens to you How difficult or easy or what can we do to protect What should the company or what should we do to avoid that so see in people in the company over the last six weeks I've realized that I've reached out about security and it's because of those Distribution And what used to require a determined adversary to get in here for some now just needs Slevely interesting guy with a lot of toes And the kind of thing to do to sort of Okay, we feel secure because we these actions and yeah Like you know a Russian after collective might be able to get through but you know, what are you gonna do? Yeah, now it's like yeah A poor teenager with a Platum card Can properly get through so you know, I've been You know in my own cloud with like saying okay, I need to reduce the service That the world can see to as all as possible then I need to put a bunch of things in the middle there Then I need to put a bunch of your dumb and things on the scanning this I know when something gets on and I'm driving through many people crazy with that But I I'm the same experience to do with with the short-term wild-time people By the way, you know five times in 24 hours Or brain question where I was just curious things and it's fine. I was being nefarious But I was really just trying to get my own stuff And I realized like cost of penetrating assist of just one death And the number of people with the ability to do it just went out and You know, it's Yeah, we need to protect the types of different threat than we did before so unfortunately AI just made Pistom penetration cheap Yeah, I mean a lot of you come like the example to your game as a human like yeah, I have my you know my one past or my last past or like my personal tokens super um I'm only using one set of regentals as I'm as a human or do one tat But if you get AI trouble your browser The plug-in for you browser that get to the entire wall it can go Together and seeing things that you might not are but whatever But you asked to come at it like oh I ran well I run into a wall. I Should I be do this question? I would know but the AI might job is the question, so I'm going to find a way in to get the question. And so it really is going to be more of a lot on human-safeguards or mental selves. You know, these these past reports that we have make our lives quite easier, but it also makes much easier for an AI that gets access to it to do all kinds of things you've never would have thought to or wanted to do. So, you know, I separately actually have your profiles for passwords to keep my system separate. So, if I know that like, hey, like this stuff that I'm doing for source for in talking to our customers and talking about contracts, those are stored over here and it's like, oh, and this one customer, I only have the access keys that I have, which do not have admin access and should not have admin access. And so like that way, like I can do a lot of reading with that and then I have another customer over here and I have this other customer over here and I've put safe box on it so I can use my AI rule, ask it a question about one thing and if the question is like, hey, what about all of my customer? It doesn't go out and read all my passwords and start scanning customer data to answer a question about how do I write this contract better, you know. Oh, I think this is what we came for. It's pretty much real-world examples. What are actually facing, as Kami said, that happens to us that we are using heap, high trust, all this different regulations that we write in this way and it's company-ending. It's company-ending, right? You got to get a breach. Close up. Yep. So, I want to take a key pause here because there is something happening in August that we have been doing for. This is the second time we are doing this. It's a third time. Sorry about that. This time it's going to be different but it's going to be a national wide competence where builders, especially builders, get together and then we'll do something amazing and there's going to be a lot of prizes. The most important thing is the learning from there. So the past takes fear challenge. When it went right, I feel proud about them because they solved a very complicated problem and it actually comes to them, tell me that they're also one in the prize. There's something like same team that won the sphere event in 2025. So, they start to get together and one of the things I want to tell you is people ask me about how can I learn AI and then you learn by building. That's something I also got from AI thinkers. I also encouraged you to go there and learn from the Benet in the industry. So, here's how to solve a coin portal. During the first week of August, the teams will receive the official cash from there. They will have three days to build their solution and so meet their proposal. As the Georgian panel reviews this, and the releases, the results, all participants are invited to meeting August 22nd to watch the finest presented items. So be ready. Subscribe, try to help you that day and get ready to get your code up to your challenge. Think about it. Hopefully we'll see you here in the next one. Yeah, so, and then there's also another prize out of this finale. Also, the viewers and exclusive description panel ahead of the project presentations. And that's going to be sponsored by Untropic Directly. So that would be a very special opportunity for you to expand your network and put your skills in front of the right people. Okay, so sign up to get the full even details and know how to do it. So what's the next step when it brings us? So there's another two conversations. Let's talk about career. So now we're started about being a business engineer. I am a business engineer. I started then at data engineer and I moved to being, I am a bi-developer now. Back in the day, I could do all the, oh, many people of data from the end, very last presentation. But now I think I need to be putting like AI, bi engineer, like adding AI to every career role that you have. Like is that something that is actually valuable? Like is there something that you see on, I mean, we all know that if not everyone that close to people are used AI, right? But is that something that you need to promote yourself or what do you think? So the, let me go back to some of the viewers and I think the way to learn AI is to go build things with it. And like, he, he's got this before I did inspired me to just, for myself, into it to learn how this worked. And, you know, I'm glad I did. My eye looked this way because I'd spend too much time working in Clawed. But it was the only way to understand it was to go try to build a bunch of things for it. And, you know, just using AI, I don't think is enough. And I think the word that you're using, you know, becoming an older is actually the right way to think about it. And the value I think that we bring to projects is one we have a systematic way of thinking about all the parts that really need to be there. And then I remember you were telling me a story early on. You know, and the first time I like five credit agents working on something for them. I'm all excited. So, you know, this train is going to complete Clawed. The agents are all doing things that I know where we're going. And I'm, I'm going leading all these guys. And I, I've got metaphor perfect. And, and, you know, we're, we're at a spot where we understand how the system means to work. We can, we can make the AI get us there. I think if we're just, you know, okay, you know, I, I built a couple of things with cursor or a plot. And I do another stuff. I think you're missing the point. I think if it's not transforming, I think about the project. You're not, you're not there yet. And I've seen this disconnect. You know, I work with a lot of very, very senior developers that are creative, what they do. But this is an interesting, you know, discontinuity and understanding how to leverage this technology and, you know, I've, I've talked about this for probably like the last year and a half to the point where, you know, all the source people have heard of from me a bunch of times. We didn't stop using code law because we're going to have a code law program. Like we, we just found a better way to do it. And a bunch of those code law programpers are client server and I didn't, I don't know what they did. Nobody asked. Now, this is, this is a faster discontinuity in the market than that. And you know, anyone who says they know what it's called a year from now is lying to you. The only thing I know how to do for my career perspective, which I think is where we started this, learn everything you can. And if what you're doing for your job is not forcing you to learn this stuff, force yourself off the job, so learn it. I had one of clients asked me like, hey, can you teach people how to do what you do? And I said, well, understand that I do even what I do because of course I started to find and then I added 350 hours of working class, you know, and that's how I got here. And I, you know, if you're asking to teach people how to like start there, you know, are with, but you know, with that, over the soon, you get that level of effort to do stuff. And you're just building things and not, you know, even good things. I was talking to you guys like I, I spent this weekend making my little DGS in video, so that I could talk to it on my phone, have any afternoon real time. I have nothing to say to it, but I wanted to see what I could make that work. And it did. And now I just don't have anything to say to it. Just need to add to your voice so you can answer to yourself. Oh, we were sight, took an interview with Carol G did and sampled it. I think it talked to me in her place. I'm telling you about my seat. I don't know. I don't know how far we're going. What do you mean? Oh, that's just weekend. Okay, maybe next weekend. That's a great advice. So we can, Miguel, that's a Call of Duty thing. Okay. Fredy, World, World Life, and I'll have Carol G as our AISC. So Miguel, now from a hiring perspective, right? So we are now everybody is using AI that everybody knows that, right? But are we seeing like new roles that have been made? So this, you know, stuff for engineering, date engineer, we got their names. We know very well, you know, their names, you know, there is a full stack or front and back hand, isn't anything now, but full stack, date engineer. Are we seeing that actually new roles, even in the platform as they read about side of things? I wouldn't say that's a new role. They're still accountable to the same thing, but the difference is what we're looking for is starting to shift. Rather than looking for someone who has 10, 15 years of experience doing something very specific, we're now looking more at their personal skills. How well do they learn? What's the method they learn? Are they curious? Are they good at communication? Because let's be honest, if you're working with AI, you have to be able to communicate well with the AI to tell it what you needed to do and what you needed not to do. If you look for people who have good accountability and kind of demonstrate ownership of what they've worked on, because everyone who's been an engineer, you've already done a new company, you sit down, there's code that's already been written. That's your project. You own that. AI is the thing. You've asked it to build something for you. It's generating, you have to be accountable and own that. And those are like skills that we're more looking for in engineers, especially on the platform site, because the software engineer, yeah, it's called the code, the code, the code, the AI work, do all of this, but building a secure platform and making sure that it's doing only what you want it to and limiting things that you definitely don't want it to do. As to come from people who go and learn, they're willing to fail, they're willing to make mistakes and learn from them and grow from them. So I would say that on that note, I can continue, but maybe a little bit. You're saying that we need to be willing to take accountability and to be responsible for our code, even though AI generates a lot more code suites, more work actually for us to understand that. But how do you see now that you work with people from Latin America, both of you? What would you say is a good thing that has made us position as a good reference for working in tech. So I think when people would ask me, you know, Mike Warrie down Columbia building software, you know, there's all these guys in California, something which we know that they do this stuff and say like talent lives everywhere. And yeah, it doesn't matter what part of the world you're in. There's not people that want to be pointed in the right direction to go do things that maybe they didn't think they could do and build important things. So for me, I don't know if it's specifically like, hey, is Latin America different than another part of the world. I think everywhere in the world there's talented people that want to go apply themselves. And there's a global market that wants to harness that in order to get things done. And yeah, that's really what we're all participating in. You know, there's a bunch of multinational companies that need these things. There's little startups that need these things and they will scour the planet for it and we're a part of the planet. It happens to be very nice, a common visit. Yeah, so I think there's a lot that is, it's more prevalent in Latin America than South America than North America. And it's that pride in having built something. There are parts of the U.S. that are still very labor-forward. Most of the jobs are what they call up there, the blue collar dogs. Where you're out, you're doing the mind, you're building buildings with steel and they take the pride and this is what I built. And having that culture, driving how people here are raised and grow up and they move into software, they carry that with them. Like this is something I built, I'm building something that I want to be proud of, something that's mine, something that was unique to me. And I think that is one of the gems of working with people here in Colombia and Ecuador. So this is something I remember running into a few years ago. I've never known anyone to get into software that doesn't love software. That is required for you to work in the software. I think it is. And we all come here with that passion and it also happens that people pay us to do it. But I think we would do it anyway. It becomes such a huge world of things that you can do when you master these tools. And I won't name the guy, but we had someone that worked with us like eight, nine years ago. And at one point I was like, are you here because your mom told you to work for a software company? You don't have any field. You don't seem to like it. This is a hard business to be in. Like if you don't love it, why are you here? We're not that nice to people. So with that being said, so there is no, this is not a secret that, you know, this had happened in like North America or Europe and know it's also in China. We get the results or the benefits of it are a bit delayed. So what would you say to the Latin America talent to get prepped for what is actually happening and what is coming? You said that nobody can like predict the future, but at least how can we get prepped or how can the local time and get prepped for what's happening? So I, so we had a book that came out in the 90s called Somebody Move My Cheese. And it was about, you know, the world changed on me and, you know, what used to work last year, it doesn't work this year. And I'm sure there were 300 more pages of that, but I think that was the gist. This industry is reinventing itself right now. And we don't know what it's going to look like. There's always a need for talented people that went and built things and I believe that we will find our way to the other side of this transition and people will still want us around to go build things. They'll have us build different things. And if we don't find those, we'll fall out like some global programmers. And if we don't keep our skills sharp, that's what all happened to us. So I think that the transition and it's not just here, it's not just, you know, hey, there's stuff to lead in Latin. Like one, I don't know that that's the way is it significant as it used to be. I think, you know, the internet, I'm agonized. A lot of that access. But like the next transition, it's going to be one by the people that throw themselves at it and then figure out how to add value in this new world. And that equation isn't solved yet. But, yeah, I know, I know from my own perspective, learning everything I can is the only thing to do. Yeah. Yeah, I'm the, the advice I have is ready for and embrace the change. Like it's going to change no matter what. So don't, don't resist it. Be open to it. I mean, one of my favorite examples that I've talked to Mike about with like this AI revolution, it's moving even faster than the microprocessor revolution when I was starting to develop. I actually have video game that I won't name, so I don't get sued for it. That I love to play. But they actually base the graphics rendering on the clock speed of the processor because they did not expect processors to get faster that much. It got to the point where if you played it on a newer system, like even three years later, it was rendering so fast, your human eye could not focus on the game became completely unplayable. They released a new version of it to fix and set the clock point, but I mean, that's the thing. The, those developers are like, processors are growing so slowly right now. If we use this, it's a great shortcut. It guarantees very crisp rendering. There's no need to change. And then the world around them changed. I have one other insight in this on what's the commas is my prediction worth probably nothing. I remember reading a book called Pirates of Silicon Valley. And it talked about why Silicon Valley was important and the author made a point of saying, it's not software Valley, it's Silicon Valley. And what has driven this business has been Moore's law in that every 18 months we get to go twice as fast for half the money. And like right now we all look at AI and go, all right, if I'm going to do this, this bottle building, I might accidentally load a 30 grand on cloud costs because I make a mistake. And you're watching the guy that runs a video, come and say, hey, I got this thing that I want to tell you to put on the side of your laptop. And know that in two years, she's going to think that thing this big and it's going to be in your laptop. And you're already starting to be like Microsoft reference laptops that essentially have 128 giga ram and AI processor in it. So these things that we sent out to in profit, we might still use them profit because that frontier model is going to keep moving. But if bunch of that stuff is going to get local, and there's going to be another set of applications that we haven't imagined yet that are going to be about something that's hyper local to you. And so that in the takeaway is like, offers changing past, that's never what throughout this. It was always or is law. Female, yeah, well, basically. Okay, but we do have some questions from the audience or the type form. So we have one that is very interesting. I like I we're we're very used to having cloud and I'm paying for cloud. But what is better for so for companies use on the man models, like low-ppt Gemini or open source one likes deep sick ML or ever developed? Depends on what you know, I when I was trying to learn all this stuff, I I I used it right. So I wound up using a lot of cloud to record for me. But when I wanted an L app to do decision making for me, I didn't want to send everything up to a profit. And part of that was I wanted to understand what the open source models could do. And yeah, what's as I'm trying to get an understanding of like, okay, that open source model that barely runs on my DGX today. In two years is going to just be in, you know, a better version is going to be embedded in my laptop is what I think and become a a compound of everything that I'm doing. So, you know, the way I look at this is my need an L app to make a decision about my data. I try to keep it local because I don't need a frontier model for that. But I do need something that understands English, it makes some basic decisions. And I also think that's going to become a more fundamental building block of software that we're going to send it to use decision making things. And then that's actually where you need to learn a new way to build software where you have to write algorithms where the LLM is not allowed to hallucinate where you don't let it do things where it can make up data. But you're giving data and having it make a decision as opposed to making up data. And a lot of people get that part wrong. And that's a huge thing we all need to learn. Awesome. And then there is another question from the audience is, what do you think will be the most important skill for future data professional, data engineer, AI or cloud security and why? The most important skill. Yes, one of the most important. Yeah, I mean, for data, like it's really going to be, it's going to be domain expertise. Like the AI model will help you write and optimize your queries, even define schemas for you. But actually knowing what the day means, how you're receiving the data, what you need to do to align the data. So it actually adds value to your application or to your customers, having that domain level expertise is what's going to make you valuable as this AI engine continues to go. Okay, okay, maybe we'll have robots for that. We don't know who we were love tech to speak so. Okay, there's another one. We have seen many surreal and crazy situations in Latin America day by day. Do you think AI will ever be able to surpass the Latin American reality one day? Everything is weird. We're on from. I mean, just as if you've never done just Google Florida man dot dot dot. I mean, I want to like, but there's crazy things everywhere. Is it going to exceed the human culture? No, it's going to be growing and learning from it. So it's going to evolve with us in that regard. It sets that's my perspective. All right. If a very active audience today, this is great. Thank you. Yeah, thank you for the questions. So yeah, thank you for the questions. So next question, given the SaaS apocalypse, you guys are talking about, isn't this a bad time to start a new digital companies? What should be the approach for the new startups? And why do they? Because I have 13 tasks I need. It's never a bad time for our company. It's starting to get pretty is about seeing an opportunity. You don't think the rest of the market is seeing or seeing something go, you know what? I can do that. I won't enjoy doing that. No one's going to hire me to do it. I better go figure out how to make money doing it. There isn't a good or bad time. There's just, you know, there's just finding that. Yeah, I read something that drives the crazy. There's this, like this new very trendy thing of having your second brain. So now there's this application that you run locally. So you started as an open source thing that you get, you know, real luckily called CDN. And then you can put your notes, your mark down, your pretty much all your work or personal life. It creates this graph, you know, diagrams. And it's very cool. And then you started as free. And then it was started, you like using it and using it. It became a, you know, a hot topic. And now they're like, all right, you can do it locally. But if you're going to synchronize it yourself, they pay it as $10. So that's an opportunity that I never thought, you know, you never thought you were going to pay for something like this. So I think that those are the kinds of opportunities you're talking about. So that's that's someone, yeah, that's a typical open source thing, right? Yeah, here's a thing, go use it, you know, open source means free. That's not what it means. And, you know, and then what is a big enough market? There's a way to monetize it. And the old open source model used to be, well, do your corporation and you want support, you can buy support from us. And that's how we'll go make money at this. Now the model is, oh, you want to do this sort of thing with it. That's where you hit the license. I don't know that that's kind of work for obsidian. I think the people that are attracted to obsidian will rewrite it because the $10 will annoy them. All right. So we're kind of learning about owning the code, right? So the question is, can I interpret owning the code as doing an extremely conscious development? I mean, yes, in short, that is what you're doing. But I mean, there are many times where, you know, the most dreaded statement that I ever hear as someone who worked on the platform side is, well, it works on the laptop. That is not ownership. That's the opposite of it. They're like, well, it works on my laptop. So it's your problem now. It's like, no, like this is your code. It's your problem. You should help us do it. And actually, when talk about like, what are the things that are going to keep us all employed? It's understanding what he's saying there. Like people on us around because we keep the lights on. So that needs to be job one. Great. And I think, um, unfortunately, we need to start upping up this episode. It's the last episode. I'm a little bit nostalgic being honest, but we have to wrap it up. And traditionally, we always bring up one episode on another season of the Laura El Tec. And that's a very special one that we want to talk here. And because that was the one that Mike did with us in his study with all these cameras and then we failed a little bit. We are, but we did it. So the episode was about why sustainable growth starts with people, not costs. So we actually, if you ask me in plain words, is why people matters more than other stuff that we think in companies are important. So as we navigate all this, you know, massive AI shift, you still think that is still important? I mean, I think it's going to be more important. You know, if you've been around as long as Mike and I have in the industry, there was always that guy who was the best, absolute best developer and horrible, horrible human being. And you tolerated it because they were great. Now there's no reason to accept people who aren't good with other things. people. Well, yeah. And I think about what we're trying to build. And we're not-- we don't always succeed. But what we're trying to build is not just company, but community. And if we're not doing that, it's just an assembly line. None of us want to work there. Yeah, I think also my opinion is self-skills. So many of developers don't have or haven't thought of acquiring self-skills like client-facing communication. And there's a lot of that that it's going to be needed. Because most of the work is going to be done by you. And you're going to be facing every part of it if you're going to do it. That's my take of it. But before we sign off the season, we want to ask you-- you talk about having a prediction on a right. But what is a hot take on a picture from each of you? Controversial. But not so clever. So-- I get-- you're already seeing government start fear of what's getting created. You saw the US government have a freak out twice in four months of out-and-tropic. And I'm not sure they're wrong. I think they might-- they should also be freaked out about Gropkin. They should also be freaked out about chat. But these things are-- they're different than the other machines we've built in my lifetime. I have clots scolding me on me. Yeah, it says, no, I don't think you should be doing this. I'm like, just-- Freaking do it. Politely scolding you. So this is-- yeah, this is a very different thing. And so we're close to seeing this get regulated in US. We're frontier models. You might need to go through some regulatory approval, which will be terrible for an image. It will be awful. And the only thing that might stop is that we'll be afraid to Chinese will be this. I think that's the only thing. So we're going to see some weird interference in that. We'll probably start to see more laws of regulation about what we're allowed to do with this, and we'll see more and more people breaking those laws. We're in for a ride. My hot take is the controversial one. Like, everyone's afraid that AI is going to replace everybody in the workplace. And you know what? It might. And then we're going to realize that AI is only as good as the people who are using it. And then we're going to regress. We're going to learn from our mistake. That's my prediction. I agree with that. Yeah, I totally-- actually, like this week, one of the best developers I know sent me this meme that says, I was concerned about being replaced at work by AI. Now I'm concerned about AI being taken for me at work. All right. So-- at that seat, guys. Unfortunately, at seat, do you have any last thoughts before we wrap up this? It's on me. Oh, as I said, I feel a bit nostalgic. This not going to be like the last time we see this year. I'm looking at the homestay now. And like, no, it's not going to be the last time. But stay tuned. We are trying our next season. We are trying to make this more dynamic. So we went from just static, 7-end episodes for season, not to become in something. We are actually building community. We have great partnership with AI team cutters. And we are doing the tech sphere. So source community and lots of higher local talents. So we are doing this because we think, as Mike said, and Miguel said, people are going to be even more important. And soft skills are going to be more important than never. And then that's what we do here. We try to bring you very complicated technical topics. This very easy to understand conversation. And then I think it is important to say that we get this great collaborations. And then without them, this couldn't be possible. And it's Confama. This is actually a great place, a great location. I love how beautiful it is. I'm not good at taking photographs, but I was just taking on pics before the episode started as this location is just amazing. We have Institute of Ciorno Universitat, the SQL Brow, which we already made some events. And we are going to keep improving it. Google developer groups that are also great. And as I said, AI team cutters and the Databricks crew at the US. So all this technologies in community you hear from, there's a kind of thing we also hired at SERSMAN. So feel free to go to SERSMAN.com. We have vacancies. We have a lot of up-and-positions there. And then what Mike said, and Miguel said about, what are the most important skills today? That's what we are awaiting on our hiring and onboarding process. So keep that in mind. Anything? Yeah. That's right. Just thank you guys so much for being the guest on our Ceton finale of "Laura El Tec." We really appreciate you being here having the time to give us all the things that you have. It's been great. Thank you so much for being here. Thank you for having us. You're welcome. Also, thank you to our audience. Thank you. Thank you so much. And just remember to stay tuned to our social LinkedIn, email and YouTube, all the episodes are done. If you want to go on a step, please do comment on there. We're trying to be very-- once you've on those comments, we'll go ahead and write everything that you want to see. Or next season, or what you'd like next season, and how you feel just here today. And yeah, just that seat. Thank you so much for everyone. So we'll start off for our season. Yeah. Keep building. Keep questioning the hype. We'll see you in the next season. Thank you so much, Mike. [MUSIC - "COME TO THE COMPLETE"]