Read the full episode transcript
Generated from the exported YouTube subtitle track. Timestamps appear roughly once per minute and link back to that point on YouTube.
You are listening to the Data Hustle podcast, where data is turned into business impact. Today, we're joined by Alexey Grigorev, founder of DataTalks.Club, creator of the Zoomcamp series, machine learning engineer, Kaggle Master, and author of three technical books. Alexey has spent more than 16 years in software and 14 years in machine learning, and today he teaches AI and data to more than 100,000 people through DataTalks.Club and his open educational programs. Before building that community, Alexey worked across software engineering, data science, and production machine learning, including several years at OLX, where he led internal machine learning platform work, MLOps capabilities, and large-scale image model systems running in production.
Along the way, he took first place at the NIPS 2017 Criteo Challenge and second at the WSDM Cup. He's the author of several books, including Machine Learning Bookcamp with Manning, TensorFlow Deep Learning Projects, and Mastering Java for Data Science. He has close to 1,200 pages of book material and several bootcamps out there. Once he finished all this stuff, he did Education with a big time capital B. The Machine Learning, Data Engineering, MLOps, LLM and AI DevTools Zoomcamps, 22 free cohorts, six courses, 41 modules, 85,000 GitHub stars as far as I can count, with the Data Engineering Zoomcamp alone at 44,478 stars as of the recording and 8,720 forks.
640 hours of public videos, 208 podcast episodes across 24 seasons with 187 guests and counting, running monthly since December 2020 with over 420 community events. If we take everything in aggregate, Alexey has been shipping one public thing every four days for six years, which to me is an incredible output. And so With all of these great things you've done for the community, Alexey, can you tell us a little bit about your story? How did you get here and how does this all happen and unfold?
First of all, thanks a lot for all the digging, 'cause I think I need to update my bio, bio across all the platforms I have, 'cause, like, what you, put together is way better than what I had. How was it like putting, something new every four days for six years? That's impressive. Okay. Anyways, yeah, your question was, how did I get here, right? so I don't know how far should I go. So I started, my education, was in informational systems, and I learned things like Delphi and, .NET, Then I started working as a Java developer. I worked as a Java dev for a few years, and then, it was pretty boring for me. And then around that time, machine learning became a popular thing, and, I took the course, from Andrew Ng, on Coursera, was about machine learning, and I thought, "Okay, like, this is what I want to do."
So I enrolled into a master's program. It was, about business intelligence. Little did I know that business intelligence have nothing to do with, machine learning. I only found while doing the master's, but still was fun two years. I ended up in Berlin, doing my master's, and I have been living here for, long time, since 2014, I think. so then I worked as a data scientist, data science, data scientist/data engineer, ML engineer. Like, my title was always data scientist, but I've did-- I've done so many things across, kind of full stack data science experience. and, I ended up, being a principal data scientist where I led a small team. We called, our team like special ops team.
Like, whenever somebody needed help with productionizing services, we would help, like kind of data forces. but also we were developing, a platform for making data scientists more effective, efficient, and let them ship data science projects, machine learning projects faster. and then at some point, I started DataTalks.Club. so it was COVID. Everyone was working from home, and, people wanted connection, right? I was one of them. And, I thought, "Okay, maybe I should also start a community." And the reason I decided to start a community was that, I was, A lot of people were reaching out to me on LinkedIn and other platforms asking for career advice, and I would, have a one-on-one consultation with them, 'cause I thought maybe this is something I wanna do, as my kind of hobby, consult people their career.
I did like around 20, 25 consultations, and I thought like, "Okay, it doesn't scale." Like, it doesn't really... Like, I cannot do this. I can do this maybe one, two, maximum three consultations per week, but then like, okay, since I have other things, like, it doesn't really scale. but how can I multiply my, how to say, multiply, my help? Like, how can I reach to more people? And I thought like, since I have helped so many people, how about I put all these people together, and when somebody else reaches for help, these people who I already helped, they can help others. And this is how, the idea of DataTalks.Club came to be. So I created a small website, just like with a email, field that you could, could get in.
I invited all the people I knew, invited all the people I consulted I put it in a few places across the internet and people started joining. And a few years after that, we had all these things you mentioned like, podcast, courses, workshops, like all these things, and it was taking so much time because like I was running, this, team at OLX and then doing this work in parallel. It was just too much, right? And then I had to, to decide, okay, what do I actually want to do? What do I want to focus on? Is it the work I'm doing as a data scientist or is it the work or my hobby I do as a, a community manager, community organizer? and then I thought, okay, like I need to figure out how do I earn money with the community.
I took a sabbatical. I figured this out. I found my first sponsor. The money I got from the sponsor, I reinvested into automating the community and offloading most of the work I was doing, and delegating it, to an assistant. Then I managed to, finally have some time to think about like how exactly I want to approach this. then the sabbatical ended, and then I continued finding sponsors for the community, and then eventually I left OLX, which was like the best place to work as a data scientist, but still, like I wanted to do something myself. So eventually I quit to focus on DataTalks.Club. And it happened three and a half years ago. So for three and a half years I've been running solo, focusing primarily on DataTalks.Club, but also there are some other things I'm doing in parallel to that.
So did I answer your question? I think it was pretty
absolutely. Yes. No, there's, there's a, I think a lot of interesting things to zoom in on. I think you, you definitely make it sound like it was all a very smooth and then, and linear journey, and I-- Yes, exactly. It never is. I think we could definitely zoom into a bunch of, of aspects of that story. But before I dig into this, right, you do mention you've been, like, consulting people with their career for quite a while now, even way before you started doing this as a hobby and as a business. So in your perspective, if somebody were to start today to create a plan for their, data science career, maybe they're like an early-stage professional, maybe a mid-level professional, what would they need to focus on to stay competitive in the next five years?
Because, I mean, as everyone knows, you know, AI is taking the scene, but other than AI, the pace of technological development has just been increasing steadily every hype cycle, right? Like, back in the day, you could learn technology and stay relevant for fif-fifteen or twenty years, and now it feels like it's maybe two. So how, how should people adapt in this environment?
So first of all, you mentioned data science. If you plan to, build a data science career, and I would stop you right there and say, "Don't build a data science career," right? So data science is, I wouldn't say like a dead thing right now, but, to me, data science, it always was a confusing term, like what exactly does it mean, right? And then at the, at the beginning it meant everything, right? But then we started to specialize, like there were data scientists, data engineers, machine learning engineers. At the beginning, data scientists were doing all that, right? but now since we have, data engineers, we have machine learning engineers, the role of a data scientist is pretty narrow, right?
And not many companies need that. And in my opinion, it's better to be a generalist rather than very focused, data scientist. So I would position myself as a data generalist or, software generalist or product engineer. So I would focus on solving problems, solving problems that users have rather than selecting a specific technology stack. And I think this, like focusing on the problems and focusing on solving these problems rather than, okay, what is the technology for that? I think this is the-- It's not the secret, but this is the way to go because technologies change, right? So 15 years ago it was Hadoop, 10 years ago it was Spark. five years ago, I, I don't know, like something else, and now everyone is talking about AI, right?
So these things change. But what doesn't change is people have problems, and we need to solve these problems. And if you have this problem-oriented approach and you focus on solving these problems rather than technologies, I think, this is the way to go. But then the question is like, how do you exactly solve these problems, right? So then you kind of need to stay relevant, And then for that, I don't have a very good answer other than, you know, just try to not do one thing for a long time, right? So try to diversify what you do. Even if it's one project, maybe within one project you can do this. If it's one project at work and you cannot do anything else, try to explore things outside.
Take part in hackathons. Take part in like, I don't know, build your own projects. So try to see what is happening there in the world and try to see, okay, there's this new technology that people are talking about. How can I use it to solve the problems? I have a problem. How can I solve it, this problem I have with this technology? And then you can understand, okay, like this thing is good for solving this problem, and then you continue, right? And then like I think like starting with a problem and then finding the solution to this problem, in the long run will help. And then of course, you need to, build these small projects to understand which tool is good for solving which problem.
No,
pretty universal, right? It's not just data related.
No, absolutely. I mean, I, I think it's in, in any career, any job you do, if you can focus on solving problems for other people, I mean, that is a Really good starting point to keep yourself relevant, because at the end of the day, they don't care how you get the job done. They care the job that they care about gets done. no, that's-- This makes a lot of sense to me. This is, definitely something we, see a lot in the data consulting space as well. it's always about the problem. What's the problem you're trying to solve for someone? I guess,
right now to, I would suggest anyone right now to maybe zoom out a little bit from like whatever tools you are working with and focusing on tools specifically to broader things like product management, like specifications, these kind of things, right? User interviews, read a few books about this, like, The Mom Test or things like that, right? So they are not engineering books, they are more like understanding the user things. Start from there and see where you go.
Oh, that's a- always timeless advice that I guess, you know... One thing, though, that I, I think is still important, and, you know, you and your team are still doing a lot of this, is the education side, right? So, what I'm very curious about on the education side specifically, and whether it is the technical or the soft skills for data professionals, I'm seeing a lot of different perspectives when it comes to the impact of AI on education, you know, as a whole, but also in the data field. You know, some people have seen, you know, less demand for educational material because people might say, "Well, I have AI to explain things to me." others would say AI is leading to weaker fundamentals in people.
They get-- become less critical. They take more of a, the answer for granted that they get from AI. Well, yet others would even say it's actually the greatest learning opportunity ever because you have, like, your personal mentor in all these different problem spaces that you can help you to get better at things. So from your experience, you know, educating so many people over, over the years, how have you seen, like, the impact of AI on the professional development and educational side of things, in, in
Well, I can start with DataTalks.Club courses because I have not seen a decrease in demand. Like we consistently get, approximately the same number of, people enrolled in our courses throughout the years, data engineering being the most popular course, and then the, the rest are roughly equally distributed. and let's, let's take the-- our machine learning course. Like if you think about this, now building a machine learning model is just open a Claude Code session and asking, "Hey, like build me a model for this," right? And it's done. Despite that, people still come and take our courses, right? And this is-- I find it pretty interesting, right?
So I'm, I'm still trying to figure out why, right? When you have an on-demand tutor that is available to you every single day, every time you need it, why people still come to watch courses. I don't have a definite answer to that. I suspect people want to learn from people. but yeah, I'm, I'm still puzzled, but this is what I see from our data from... Like AI have been around for what? Like four years, right? You could already use ChatGPT a while back, and it was already giving you pretty good answers, right? Of course, like things are improving, but fundamentally it hasn't changed. Okay, now we have coding agents. We didn't really have like these good coding agents before.
But education-wise, like if I had a question about machine learning, fundamentally the models, the, the, the answers the models, the current models give now are not fundamentally very different from the answers I was getting like three years ago, right? And still people prefer to take our courses. And also like, okay, our courses are free, but I also teach, I have paid courses, and people are taking paid courses too because, I have an idea, I have a suspicion that when you talk to a model or when you talk to LLM, it can overwhelm you with choices, right? So if you ask ChatGPT, any question, it will give you like a five pages, result, and then you really have to force it to, like, "Hey, like focus on my question, please." Right?
So you kind of have to know how to handle it. So perhaps people just want to have like a structured plan that works that somebody thought through. And you see like testimonials that people confirm, people who took these courses, they confirm that, okay, it worked for them, and they just go for these courses because they know that it's already a journey that somebody went through. It's a journey somebody took care of thinking about, and then you don't really need to explore the unknown, right? So the-- I guess like when you're just entering, a field, it's way easier than trying to, navigate this unknown yourself. So this is my suspicion. yes, and people ask, ask me like, "Hey, like how is your business affected?" But it's not.
Like I'm happy, like I'm keeping my fingers crossed that it, will not change. but this is the situation I see right now.
That is very fascinating indeed. I-- That is i- indeed quite puzzling, because I do see quite different answers depending on who I ask. And, you know, as you indicated, like, for your business, it's been pretty steady and, it seems that people do appreciate, like, that structured journey. Would you say that's perhaps, because, you know, some of the course material might be focused on people getting started in the field, and they might have, like, unknown unknowns, and they don't really know what questions to ask, so they wouldn't even be able to use the sort of super powered tutor that well, and they prefer having somebody, as you say, like guide the journey.
Like, I guess something interesting in that, right? Like it, it's not really about the knowledge. It's, it's more about how do you structure a learning journey in a way that it's accessible to someone.
Yes. I, I think, yeah, you-- this is the, the right, like this is the, the reason, right? because like the way I learn now, and you probably too, because we already have some background, some knowledge, some fundamentals in a particular area, and if we want to learn a new technology in this area, we already have the building blocks, and we understand, okay, like this belongs, this puzzle belongs to this area of our, puzzle, right? So this piece belongs to this area of our puzzle. And then, we can ask AI, and then AI will tell us, and then because we have the fundamentals, we will know what it mean and how to piece these things together. But for somebody for who like, I don't know, a word data platform doesn't mean much, and then all the individual parts of the data platform also don't mean much, then for them it will be difficult to understand that, okay, now we have, I don't know, these Snowflake things that replace Spark, right?
So because, they haven't seen this, and if somebody just explains what a data platform is and what are the components of this data platform and how you put them together to solve a problem that you have, this is what we do in the Data Engineering Zoomcamp course, right? We go from like raw data to a dashboard that is running, and then we build all the things around that to make, the, the journey from raw data to insights possible. for somebody who's outside of this or maybe like tangentially working with that, I think this is just very helpful to just see all the steps laid out. And also I need to thank actually AI, because, not only it's not harming what we do in DataTalks.Club, but it's also bringing users.
And right now, if I look at the traffic sources, so many people are coming from ChatGPT because ChatGPT is recommending, our courses. And not only free courses, like this, course that I have about AI engineering, it's a paid course, and it's also recommending that course. Imagine. Like I just have to thank AI for its existence because it's actually not only helping people to, you know, start a career, but also points to, our courses that we do.
That's a very good point. You guys probably have some good AEO or AI engine optimization there going
I wish like it's, it wasn't really, how, how to say? Like I didn't really do this knowingly, but just because, DataTalks.Club, DataTalks.Club has been around for quite some time. If you open Reddit and you put something like Data Engineering Zoomcamp, you see a lot of discussions about it. Some people hate it, some people love it, but there is a kind of digital footprint about the course, right? And, for these systems, my understanding is if they see these public discussions, it gives a very good, strong signal to the model that this is something they need to recommend. Because people ask questions on Reddit, people get answers on Reddit and similar platforms, right?
So, because these models are kinda, taught to answer questions, these platforms like Reddit, they give a lot of, information to the models, right? And since people-- Like it wasn't something I was doing, it was just people were doing, that themself. we just got in a, in, in a lucky position to, to be now where AI recommends, our courses.
It is funny. I, I, I immediately think about the early days of, GPT. I think GPT-3, their initial training dataset for, much of their, you know, early knowledge was Reddit, and they would just do some very simple filters. Like, if there's three upvotes, we consider this a signal. We will include this in our, in our data. So I, I guess what we're saying here is if, if you have a course to sell, make sure people are talking about you on Reddit. Still relevant to this day.
Yeah. Well, now I don't think Reddit is very happy about, that though. Like, so I, I would ask people to still be people and not ask, you know, OpenClaw to just go and spam on, on Reddit. Nobody will benefit from that. So please, like if you really want to promote your course, be a part of a community, right? So don't go and spam, but join a community, see what people are asking about, contribute to the discussions, right? And sometimes if whatever you're selling is relevant to the discussion, then you can bring it up, right? So, but make it organic. Like, and, because I manage a community, I hate when people join and start spamming, right? So please, please don't do this.
Like, yes, Reddit and other communities is a very good source, but handle it with care, right? So handle it responsibly and with respect, right? So don't just ask your OpenClaw to go and spam the community. Because like for me as a community manager, this is such a, not waste of time, but like I have to deal with these bots like every, every day, and I have to ban like three to five people daily, sometimes even more, and this is just annoying. So please don't do this.
I, I'm so sorry you have to go through that. I guess that's the, that's the curse of a big community, right? The, the spam control of AI nowadays is,
I think in worse 'cause at least in Slack people have to go through registration process, right? not that, AI cannot do this. It- passes all these steps easily. But at least there is some, I don't know, gate. While with Reddit, I think it's like more open.
we know the journey takes a long time to get-- to build a big community or a business or all these things that people want to do with their, you know, data knowledge. How much time, effort, dedication, money and stress does it take to turn your knowledge into something that's impacting a lot of people, whether for free or paid? And what does it take to, like, get to a point where it's sustainable and y-you can do this as a full-time job?
yeah, I, I don't really have an answer to that. I wasn't keeping track. like when I started the community, I was mostly spending money without earning money. but more importantly, I was spending time, right? And then this is something you cannot really quantify. But I was just having so much fun. Like, it was just giving me so much back. Like for the time I invested, especially when we started running courses and people were thanking me for doing these courses. People were writing me how the course changed their career. And it just kept me so motivated, and it kept me going. So I think this motivation is the important thing, right? So then if you have motivation, then you have the persistence, you have the commitment, and this is another thing.
In COVID times, many communities were starting, right? So the inspiration for me to start a community didn't come out of nowhere, right? 'Cause I saw other creators, other people, many companies, they were starting the community. And if you look at this now, how many of them survived? Not so many, And I think what is the difference between what I did and, for example, there is a MLOps community. They also started during COVID time. They are still here. a few other communities, they are still around, like, yeah. So I think what is the different-- what is different between them is the consistency, right? So no matter what, like you just, you have to do this, right?
And okay, yes, sometimes you see like there is, there are ups and downs, and these downs are demotivating, right? So you host an event and nobody shows up, right? It is very demotivating. But it doesn't mean you should just, you know, completely quit and never come back. But it, it just helps to, to be consistent, to continue. And when you get good feedback from people, it helps you be consistent and go through these downs, and lows, right? And when, It's always like that. Sometimes you have, a lot of interest, sometimes it goes to more like a low point. Then it goes up back, goes back up, right? And then, if you have motivation, you just need to patiently continue doing what you do.
And then, yes, sometimes you need to spend money. but then at some point, like I had to, I think I wa-was telling you to, to make it, profitable, right? And then it requires some planning. It was very fun. Like for me, it was like exploring the unknown because I have, I had been an engineer my entire life. I never launched anything. Like I didn't really earn money online. So for me, listening to all these podcasts, reading all these books, it was so interesting. So I just wanted to do this. I wanted to continue. And, I think this is what kept me going. So I think this is the main, I wouldn't call it the ingredient, for succe-suc-success, but, at least for me, this is what was, the main factor, that, okay, the community is still running.
We're still doing these things, and not only we're running, but we're expanding the courses we're doing. we are adding more courses to the catalog. I'm going beyond DataTalks.Club to do many other projects, and I think that's the main thing that keeps me going.
So in summary, you're saying you have to do things that are for a greater purpose than just yourself and do them consistently. Is, is that what you're getting at here?
I guess, yeah. So consistency, this is what matter, this is what matters. But also, like at the beginning you started like, why do people give a lot of value for free, right? where does this come from? And, Why did I do this? Right? And at the beginning, like I wasn't really earning anything with DataTalks.Club, and I wasn't-- Like for me, it was never the plan. I just wanted to have a place for people to hang out together and never thought of actually monetizing it from the beginning, right? So this came later when it started taking a lot of time. And for me, like this giving away a lot of value for free is, is just such a career multiplier.
So for me, I always was a part of online communities. Like when I just started my career as a Java developer, it was like two thousand and ten, the first thing I did when I became-- I, I got officially hired as Java developer, I joined a Java forum. And I would see what kind of questions people ask, and even if I don't understand the answer, I would try to do some research. I don't understand the question. I don't know how to answer the question. I would do some research, and then with the knowledge I collect, I would try to answer the question. And that not only I helped the, the person, but also I learned something new. And that over the years, it accumulated.
so then like for me, passing interviews was very simple. People knew me. People could refer me. I could get like so many job opportunities. I relocated to Europe from Russia. then I found, like when I started being interested in machine learning, I did the same thing, right? So I joined a few online communities, and that led to so many good things, right? And this knowing that, okay, you're sharing something and just for the pure joy of doing this, of helping somebody, but then also at some point the, the reward comes back, And the rewards are usually like the opportunities you wouldn't otherwise find, the connections you wouldn't other-otherwise have.
And just knowing people, other people in this field, it's always good, right?
Yeah, so really you've
like why people... Yeah.
curiosity and, just en- really enjoying helping people out and, you know, learning all the things along the way whi-while doing it. That's-- And I'm, I'm just hearing a genuinely curious person that, you know, loves to be helpful to other people. And I guess what you were kind of indicating here is if you do that a lot, then, you know, there is, over time, it pays back through opportunities, through referrals, through, as you indicate, easier job interviews, a bigger community. But it all comes from this place of, of, of, of genuine care and passion first.
Yes. And passion, probably being more important thing, right? 'Cause like if you don't have this, if you are not-- if you don't really care, and you're just there for the money, which nothing wrong with this, right? But then probably you wouldn't go, to the extent that you are coming after work and doing like personal projects, right? Again, there's nothing wrong with nine to five. For some people, this is the answer. For me, it wasn't, right? So I want-- also always wanted to do something extra, and this extra would be hanging out in forums, online communities, writing blog posts, writing books, creating courses. And it contributed so much to my career that like it had such a snowball effect, and it just keeps multiplying, keeps growing.
I do wonder, I mean, I hear this, this version of, you know, of, of a career a lot from people like yourself that have built like very big public things. you know, over like all the students that you've had, all the people you've had on the podcast, all the people you've worked with, what have you found like distinguishes those that have made, you know, a lot of progression towards the career goals they care about? Doesn't need to be financial, right? But just what is important to them. is it, you know, people like yourself or do you see other patterns in there as well?
well, you know that there is this thing called selection bias, right? So in my answer, I have this selection bias because like the people who I, hang out with, the reason we got connected in the first place was that, let's say they joined the community to do the course and they stayed in the community because they liked the, the atmosphere, right? And then we're still in touch, so I know how their career is progressing. I know what they do. But the reason I know this is because we kinda click on the same, values, if I say so. and that's why like I can, I can follow their career because of that, right? But there are so many people who I did not stay in touch for whatever reason.
we didn't cross, our path didn't cross or crossed, but only like, I don't know, maybe they did a course and then they left, and I don't know what happens to them. So I don't really know. But these things that we were talking about, these things that we are talking about, they work. Whether something else works, I think it does. Like being a good specialist and really knowing your thing well and, helping the user being pr- you know, problem oriented, even if it's just nine to five and you have family and you spend time with your family without, you know, hanging out in online communities, maybe you have hobby, there's nothing wrong with that. And I think these people can still be successful.
I don't want to like, you know, downsay this. It's just for me, the path I took, and I see other people who took the same path, and I see what they did, and it works, but probably other things work too. Right. So I, I, I, I'm not sure if this is the answer you wanted to hear, but, yeah, with the data I have, this is the only answer I can give.
I think that's a very i-i-interesting answer. I, I don't think I've heard this one a lot. I mean, it's, I think a lot of us get caught up with, you know, the, the, the w- the way and the things that have made us successful and the people around us that we know, that we, like, emulate or, you know, put on a pedestal that we want to, you know, celebrate. I, I think this is a very interesting answer, and I, I think it So boils down to what you mentioned.
one guy. Yeah, so I worked with one guy. I mean, I worked with so many people, but one guy I really remember. I worked with him on the same team, and he was doing really amazing job. But he would not go to meetups. He would not, take any extra courses. He would just come to work, do his work, do his, do his work really well, go home, spend time with his wife, then next day come to work, do this over and over again. And then at some point, we were choosing who to promote and he was like such a good, team member. And he said, "I don't want a promotion. Like, I'm really happy with what I do now as a senior, and I don't want to have new responsibilities because like, I like what I do now.
I like my scope. I like my, my impact, and this is where I want to stay." And like, that's why I still remember him, right? So many people are like me, "Oh, I want to get a promotion. I want to like try having more responsibilities." But he was calm. Like he, he is from Asia. Maybe that's why like they have this zen that I don't have. but he was like, you know, just chill, doing nine to five, doing it really well and being really success-successful, right? And without like being overly too ambitious, right? So he didn't want to be elite. He just wanted to be a senior data scientist, has, have his impact and do his job well and continue doing this.
And I think there's nothing wrong with that. I actually admire these people, right? Because they are of their consistency, right? was it, Bruce Lee who said like, "Don't fear a man who practices ten thousand different, punches," right? "Fear the one who practices one ten thousand times."
that is indeed Bruce Lee, I think. I've, I've, I've kn- I've known people like this as well in my personal career. I've always had, quite a bit of respect, like yourself, of, you know, well, when you're able to say, "This is what I want. This is enough for me. I'm happy." It is, it is so hard to do that in an environment that's always pushing you to say, "No, it's not enough. You need to do more, more things." But talking about more things, though, I think that kind of, leads into this question I have here, which is, I was looking a little bit in your public commit history, and just as a comparison, in 2021, you made 360 public commits as far as I can get from, from, from GitHub.
in the first months of 2026, it's been 14,698, which, you know, for some people is, is quite a big of a jump. it's about 66 a day. I'm very curious, you know, how has vibe coding and AI, a- and AI, like, enabled you to do more things, and what are you doing with it, like, to run your business and, a-a-and your empire, so to speak?
Empire. Okay. That's a very ambitious term. I mean, yeah, like to be fair, twenty twenty-one, so I was working as a lead, data scientist, so I wasn't really doing much coding. And then, as, my, well, my personal project, my hobby project, I was running DataTalks.Club. And at that point, DataTalks.Club was not very, Like I was doing a lot of work manually, so there was no coding at all. Like, so organizing events, reaching out to, potential speakers, like it doesn't require any coding. Like I, I did need to put together a website. It was super simple, and this is this three hundred commits, but like I didn't really need to do much. but if you look at twenty twenty-two, I think it's like, of course, it's not like tens of thousands, but it's more.
So then I started to like, okay, I have this, thing that I'm running. Now I need to automate it. How do I do this? At work, I wasn't doing a lot of, hands-on stuff. I was more like, I also didn't use Gi- GitHub anyways, so we had like GitLab, private GitLab, so my commits wouldn't show. but yeah, so that, that's why like the number maybe is pretty low, but I was still doing a lot of things. They're just not code commits, right? right now I'm trying to automate everything, right? So if there is-- If, if there is a thing I need to do, I try to automate. Like, the way I run DataTalks.Club has changed because now, agents can do many things and, I have two people on my team, so we are trying to look how to make their life easier too.
And so for example, one of them is a community manager and assistant, and this person has to do a lot of manual work. And I understand that this is very boring, so I'm looking for ways to how to actually automate it, with AI. And, every time now I have an idea because I have, three subscriptions. Like I have, Claude, Claude, what is the twenty X Max? Like I have Codex, the Max one, then I have Z.ai subscription, right? So I can test any idea very quickly, I can just, I have an app that I built also that works on my phone So I can open this up, I can dictate my idea, and a team of agents would start building it, right? So then testing things becomes so cheap and so fast, and most of these projects don't see the light, they, they die, right?
So there is a graveyard of projects, but some do. some I continue working on. I am rebuilding the DataTalks.Club website right now as we speak. well, this, today I ran out of tokens, so I'll have to continue on Thursday. But, the-- like I ran out of tokens for all three providers, can you imagine? Claude Code, because I have, a workshop today, so I had to leave a little bit of usage, quota for today. But then I'm basically like, I have to go touch grass for the, the rest of today and tomorrow, I guess, which is also good. But yeah, it's, it's mostly like the-- trying things is very cheap now. and I just keep these things, just in case in my GitHub repo.
So some of them I delete, some of them I keep. You see this number of commits. But some projects, they stick around. So for example, I mentioned I have this app for talking to agents, right? It's been like a few thousand commits just for this app. then, for this DataTalks.Club, also already like, I don't know, it's five hundred commits. because like it's very iterative proj- p- process, right? And if I have like hundred percent usage in Claude Code, I can do like I can work on nonstop for a couple of days, by I, I mean, I can instruct my agents to work because I always have backlog. I think this is another thing, right? So I have a backlog. That's why when I have backlog, I can just take-- tell agents, "Hey, go and work through this backlog." And, this is something, by the way, I teach in one of our courses, AI, AI Dev Tools Zoomcamp, where I explain how you can actually configure your coding agents, how you can set up the process, how can you have a team of agents, so
then you can turn any idea into reality, And the course starts in kind of the month. Anyways, if you just Google AI Dev Tools Zoomcamp, you'll find, you, you'll find
We'll make sure that, So, yeah. make sure can find all that. Yes.
Yeah. but yeah, so just because now code is cheap, commits are cheap, I mean, by cheap I mean two hundred, per month, then multiply it by the number of subscriptions you have. But like time-wise, it's cheap, In, in terms of, you know, money you spend, now I'm looking at optimizing the overall spend. So what I'm doing right now is also, yesterday we had a podcast, right? And then I told you that I vibe coded a replacement for Slido. For Slido, I pay two hundred per year, right? And two hund- hundred per year is one month of Claude Code subscription, right? So I'm thinking, okay, if I cancel this now, and if now the-- I rewrite it, I vibe code the whole thing, and it costs one dollar to-- per month to run it on my own, cloud, right?
I mean, my own account, then it's one month of Claude Code for free, right? So why not, take this opportunity? So then by now I replaced, Slido that we used yesterday. So, and this replacement that I wrote, I tested it on already on two events. It works fine. then, okay, I replaced Zapier. Zapier is like a tool for automating It's not exactly like drag and drop and connect like Zapier, but I can just tell my agent, "Hey, like I want when X happens, I want to have Y as an action." And because of the framework I have, it can easily happen, right? So then I canceled Zapier, which is like two hundred and forty per year, I believe. Then I canceled Airtable because it's just a table, right?
So then because now it's so easy to have a form with pretty much everything, you don't need Airtable for keeping your data, right? So then I canceled Airtable, and now I have a very big project of rewriting Mailchimp, and right now I pay like around eight hundred per month for Mailchimp because we have like pretty, pretty big, mailing base, mailing list. so then if I run it on AWS, it would cost around like one hundred, one hundred and fifty euros per month, mostly for sending emails, right? So this is where the cost would be, plus maybe fifty, dollars, or euros for the, the actual compute, right? For the actual service, which is like four times reduction, right?
So like these things, I just, okay, I have an idea, I dictate this idea, I see how it plays out, and if it's viab- viable, if I can use it, I keep it, right? And that's you see this, as a result in, in my commit history. I think I started doing this like in September last year. before that it was like, more chill. I was more copying code from, ChatGPT to, my VS Code or whatever. I was already experimenting with Copilot, but then I started to come across all these coding agents where you can just get an idea, s- give them an idea, and go drink coffee and come back. at the beginning, the code wasn't great, but because the code wasn't great, I started to invest in the processes, right?
So the code isn't great, but like how can I make sure that iteratively we can improve it? And this really helped, and this is what you can now learn from me, like all this experience put together in this AI Dev Tools course.
I can definitely vouch for the phone and dictating to AI agents. It's been, unbelievably addicting. It's almost hard to stop coding at some point because you can do it all the
it's like a slot machine, right? So I, I think like, it's very addictive, maybe not in a good way too, right? So you pull the lever, and then you get a reward immediately. Which wants you to pull the, to keep doing this, right? To keep telling agents to do work. And when you run out of tokens, you get this, how do you call it? I don't remember the terms, but like you kind of start, like start being agitated. I don't know the English term for that, but, you know what I'm talking you don't get yes. Exactly, yes. Yeah. So you don't get your drug and you're like, "Okay, when can I get it? Like two days, I cannot wait." this is probably concerning, but maybe like, we'll find a way to deal with this, I hope, right?
Or maybe in a year or two we'll have this, you know, vibe coders anonymous groups. Well, now we're but like, you know, maybe these things will get serious.
Hey, I mean, look, I'm just gonna close my eyes and pretend everything's just all rainbows and butterflies. And for those other people that have a unbelievable addiction to AI and AI tokens, where can they find you and your bootcamps if they wanna do this and other stuff with you?
Find me? I mean, if you just put my name in Google, Alexey Grigorev, you will find me. but also LinkedIn and Twitter are probably the best places. And lastly, I have a Substack where I regularly post. So probably this is the best, way to, actually make sure you get what I write about every year in your mailbox, because with social media you follow me and that doesn't mean you will actually see my posts, right? But with Substack, if you follow me there, you will get an email every week in your, inbox. So probably these, three places are the best ones. But also, like if you just Google my name, you'll find all my contact information.
We'll make sure that everyone can find all the links. I definitely have quite a few. and I would like to thank you again for joining the podcast. This was a very fun episode, honestly. I, I really enjoyed this conversation.
Yeah, thanks. Thanks for having me. I'm gonna stop
