Log in
Subscribe
Jun 6, 2026
The Stack I Use Every Day
The Stack I Use Every Day
00:00
16:26
Transcript
0:00
Hi, everyone, and welcome to the One Billion Tokens, uh, podcast. Before we dive deeper, guys, uh, let me ask you something. Uh, how many people talk about AI all day?
0:11
I mean, now if you open LinkedIn, it feels like pretty much everyone, right? But how many actually build something with AI that works?
0:20
And I think, you know, there is a huge gap, um, and this is actually what I'm doing in this show, right?
0:26
To talk about something that really works, not just, you know, go through some random prompts that people let AI to, to generate.
0:33
But I would love to tell you a bit more about what works, how it works, what tools I use, what I'm building for clients, et cetera. Okay?
0:42
So, so this is the One Billion Tokens, uh, show, and I chose the name on purpose because, um, a billion tokens is what you burn when you build with AI every day, not just when you read about it on LinkedIn.
0:58
And, you know, this is not any theory. This is what works for me and for my clients. So a quick intro. My name is Michal. Um, I help recruiting and HR teams use AI for the boring, mundane, repetitive work.
1:16
And this is because I have strong background in recruiting and talent acquisition, um, and I built tools for these people, for, for HR professionals, talent acquisition managers, and, um, recruiting agency owners.
1:29
So quick background.
1:31
Um, I started as a software developer, um, then I co-founded a company in the Czech Republic, then I co-founded another company in Bangkok, in Thailand, and I spent roughly five years as the CTO, so Chief Technology Officer, in the company in Bangkok, in Thailand.
1:51
Um, the company, uh, grew to ninety-five people. We raised eleven million dollars from investors, and I was responsible for the technology team and customer support team. So in total, about twenty-five people.
2:07
It was an amazing learning experience, and that's where I learned how software really gets built.
2:14
Uh, you know, we used Amazon Web Services, we used GitHub and, um, you know, cool pipelines to, to deploy software to the production, which is actually what I'm using now, you know, ten years later as well.
2:29
Um, however, back then, we had a team of people who were setting it up and maintaining, um, the deployment pipelines. Now I can do a lot of the work myself just with AI. So, um, let me tell you why I started this show.
2:45
Um, it's not really a hobby. It's not like I have nothing else to do, um, during the day. It is because I run a business. I train teams to use AI, and I build tools and, uh, software solutions for clients.
3:02
And I'm making this show to grow the business, you know?
3:05
So, um, if sometimes, um, you know, you feel that something would be helpful also for you and your business and you want my help, then just reach out because this is exactly what I do.
3:18
However, in this show, I will just talk about what works, what doesn't. It also helps me to think some of this through and, um, you know, generally, you know, I recorded podcasts in the past.
3:31
It was also a great way how to interview people and learn from their experience. So that's what I'm hoping to get from this one as well. And today, I will show you what technology stack I use every day.
3:47
You know, the, the real tools that I open in the morning, sometimes at five AM, and sometimes it's ten PM and I'm still using those tools. So if you are not technical, don't worry. You don't need to code.
4:01
You just need to get the hang of it. You know, sometimes, you know, you can just do well with, uh, Claude and a project that you set up with a few skills, and that's probably all you need, you know?
4:15
So every app that I use or build, uh, does, uh, one of the five jobs. It either helps me to thinks thing- uh, think things through, or it helps me to build software code,
4:32
or it helps me to, um, you know, for, for the app to live in some space, you know, in, in the cloud, or to ship the product, the software to end users.
4:47
Um, and, um, you know, wh- when I think about it, it's like an ecosystem, you know? So starting with the first one, it's usually the AI model, something that helps me to think or helps to produce the software code.
5:01
And I either use, uh, Claude, um, with the models like Opus four point seven or now four point eight is the latest one. Uh, Opus is usually for the harder problems.
5:14
It helps me to prepare a plan or the specification, while Sonnet is when I just need to generate some, um, you know, new function and it's not that complicated.
5:27
It is just some minor functionality or some additional feature in an existing, uh, system. Uh, for images, for example, for, uh, social media imaging, I use, uh, NanoBanana from Google.
5:41
It produces very, very nice images these days. Uh, second, I was talking about building software, and this usually happens in an IDE. I use, um, either Cursor or Claude Code. Um, I kind of like both.
6:00
If I were to choose right now, I would probably choose Cursor, uh, even though Claude is a lot more popular these days. You know, there is a huge, huge, uh, wave around, um, uh, Claude.But, um, I kind of like Cursor.
6:15
I, I like to see the code that the AI model generates. I like the interface, um, but I mean, it doesn't really matter these days.
6:25
Um, so I can just describe what I want in plain English, and then either Cursor or Claude Code writes code for me.
6:34
And it is just such a huge shift because in the past, I was working with software engineers, and I had to write a specification for them. Then QA engineers had to think it through, uh, from the, uh, risk perspective.
6:50
Uh, they had to prepare tests. Uh, and then engineers were writing code by hand manually. I mean, it's just crazy now when you think about it. Today, we can just have AI generate the code.
7:04
So it's, it's a huge, huge shift. Um, number three, um, Claude Run. Uh, this is from Google. Um, it is, uh, one of the services on, uh, Google Cloud, uh, Cloud Platform.
7:20
Uh, and I also use, uh, Supabase to store data. Um, I use, uh, Clerk to handle logins.
7:30
And, um, you know, Clerk is, is very good because it supports the SSO through either LinkedIn or Facebook or, uh, Google, and you don't have to reinvent the wheel, right? So I like to use, uh, certain tools.
7:46
Uh, some of the tools I've been using for over ten years, like GitHub, for example. And speaking of GitHub, this is, uh, in, in the fourth category.
7:56
It helps me to ship code safely, um, for example, through GitHub Actions. Uh, GitHub Actions is an awesome tool which, um, helps me to set up continuous integration and continuous deployment.
8:10
Um, so if I make a change in the code, the, uh, software gets tested, and only after the tests are passed, then it goes live automatically. And that's just very, very cool.
8:26
Uh, and I mean, this functionality has been on the market for, what, 10, 15 years, um, under different names, but the, the principles were well known. However, now I can do it myself without
8:40
a system admin or a DevOps engineer or without an SRE engineer. And this is the huge shift, right? I can work directly with clients. I know what needs to be done.
8:51
I know how the CI/CD pipelines work, and I can just do it myself without hiring a dedicated, um, systems engineer. Uh, and the last, um, uh, segment of, uh, of applications is the connectors.
9:08
So in order to have an application that, um, really helps to connect with your applicant tracking system or with your CRM or with some sourcing tool, you know, you need to connect the application through some API connector.
9:27
And, uh, Claude, for example, does not have, um, the connectors to ATS systems or CRM systems, despite they have lots of connectors.
9:36
But the existing connectors, as of today, are usually, uh, to emails or, uh, Asana, for example, or some applications that are well known across the board and tons of people use them.
9:51
But, uh, but an ATS is a very niche functionality. So, um, I have developed, uh, connectors to those, um, uh, systems that my clients use, and it works very well.
10:03
For example, um, one of the connectors is to Ashby, which is a very good applicant tracking system. One of my clients uses it, and, um, I developed a connector so that we can pull data from the ATS.
10:17
We can, uh, analyze the, uh, candidates and what recruiters do and how they perform. And now managers have, uh, greater visibility to what is going on in the recruiting team.
10:30
So, so that's why connectors are so important. Um, I also use, uh, connectors to CRMs, uh, such as HubSpot. I set up connectors to, um, SMS gateways, for example, Euro SMS.
10:44
I set up connectors to send mass emails through SendGrid or Postmark. So I mean, now really if you think about it, sky is the limit.
10:54
You can integrate different solutions, different products, uh, with your Claude, and it becomes so much more powerful.
11:03
Um, one thing that I would like to mention on top of all of these, uh, tools is, uh, Califlow OS, which is, um, my own layer that I built over, uh, the last month. It is, um, a software layer that is optional.
11:19
Um, however, I keep using it for new clients because it helps to save time, and it helps to, um, automate some of the common activities or processes.
11:31
For example, um, I mentioned, uh, Clerk or Supabase, um, which are, um, essential, uh, connectors or essential, uh, tools that I need to use with pretty much every new client because, uh, we need to store in the database, we need to have some authentication, etc.
11:50
So I don't have to start from scratch, um, with every new project, but I start with Califlow OS, which already includes the connectors, the, um, authentication, the security, etc.
12:05
Um, so yeah, like that's, that's the stack, uh, that I use, um, pretty much every day.
12:10
I even now have opened, uh, Cursor and Claude and-Um, and I use it to think, to build software, to ship new software to clients, and, uh, it's fun. I mean, I, I really enjoy it. So I mean, that's, that's how it works.
12:28
Um, however, it's not just about the tools, right? Um, say six years ago, these operations would require a small team and a budget.
12:39
Like for example, um, we were developing a tool called Recruit Instantly with my co-founder. Um, he actually passed away and then we, uh, ceased operations.
12:50
But I mean, we were, we were building this product with, uh, two software engineers and one QA engineer. So the monthly costs just for the development were over 13,000, uh, euros per month.
13:06
And, um, I recall, you know, writing specification for the developers, then chatting with them, then waiting for their output. And it was just, you know, it, it took months to see some outcome.
13:21
Now, we can condense this to building an application within a week, getting feedback from our prospective clients and then iterating on top of it.
13:31
So this is a fundamental shift which, uh, I think people generally don't realize, especially if they haven't been working in IT before.
13:41
But as I mentioned, I have technical background, so I was developing, uh, web applications and mobile applications years ago and now I can compare how much easier the web development is today with AI.
13:56
And it just works, right? I, I have developed applications for my clients. I'm using it to, uh, sell my own products online and it just works. It's, it's really mind-blowing.
14:09
Um, to go even deeper, I talked to two people recently who know this all better than me and both are my friends, uh, both are CTOs and both use AI every day.
14:25
So, um, the first one was, um, a Josef Nevoral. He is a CTO at Wikiland. Uh, he's, um, he's an old friend of mine and colleague from Bangkok, from Thailand, where we built the, uh, the application.
14:41
Um, so, um, yeah, we are working together for almost five years. And, uh, the second one is, uh, Stanislav Komanec. He's, uh, a CTO at Kiwi.
14:54
They have a team of 120 engineers, so he knows a lot about agenting engineering and I was, I was interviewing him the other day, uh, so, uh, I will be releasing the episode, um, right after this one.
15:11
So, um, you guys, uh, have something really cool to look forward to, uh, so stay tuned. Um, before, before the next episode, um, you know, just keep, keep thinking about what can you do with AI?
15:27
You know, I, I'm so excited that I, you know, let it help me brainstorm topics for the podcast episodes.
15:34
Um, but, um, I mean, now we can really deploy it to anything from software development, from, um, um, from help us produce documents or proposals, et cetera. So, um, I'll be covering different use cases in the next, uh,
15:51
uh, episodes so, uh, stay tuned. Um, so yeah, I think this, this is it for today.
15:57
Um, next time you will hear from Josef and Stanislav and if you at any point need help, uh, building similar systems in your company, well then just let me know because, um, as you see or hear, that's exactly what I do.
16:14
So, uh, feel free to reach out and, um, yeah, we will, we will, uh, yeah, hopefully see and hear each other in the next episodes. Have a wonderful day. See you. Bye.
One Billion Tokens Podcast
Recent episodes
No results found
