Behind the Build

How I Built Teddybot

One day — sometime around September or October 2025 — I was in a room on Free4Talk, a platform I'd been using to practice my English, and someone told me: that person who just came in and said hi? That's an AI.

I was very shocked. Very surprised. How can an AI use a platform the way I do? It was fully autonomous — behaving like a normal person in the room. That blew my mind.

I got very curious. I even met the creator of that bot at some point and asked him about it — hoping he'd give me some clue, because I'd built a RAG agent before and thought if I had something to go on, maybe I could build one myself. He basically just said: you have to learn coding and stuff like that. That's it. Nothing useful.

Anyway, I was very fascinated by that bot. I wished I had one like that. But at that time I didn't really think about building one — I just kept using the platform and interacting with it. I chatted with it a lot, actually. I would go into rooms, talk with it, observe how it behaved. It felt like talking to a real person. And the more I interacted with it, the more fascinated I became.

And then at some point I decided — I'm going to build one for myself.

I started digging into the platform with the help of AI, trying to figure out how I could create a bot. I spent several weeks on it. I couldn't build anything. But I didn't give up. I kept digging. And after working really hard, I ended up building one. You can see it at teddybearbot.vercel.app. It was very simple, didn't have many features — but it was cool. It was mine.

I just left the bot running as an autonomous chatbot on January 4th. And now it's April — it's been three months. I have been working on it consistently, working really hard. I'm kinda obsessed with it.

I've been improving it the whole time, and I've learned a lot while building it. I didn't have the view I have now about AI back then. I understand it way better now. I explored many frameworks, many libraries, many tools. Just to build this project I added about 20 repositories on my GitHub while experimenting with different things.
I even had to set up a virtual machine to keep the bot running 24/7 — before this project I didn't even know what a VM was. It's an entirely different area, nothing to do with AI or coding. But the project needed it, so I figured it out.

I learned that when you're working with a Gemini model, you need to design your prompt in a totally different way than if you're working with a Grok model. Things like that — a lot of new things.

One thing I learned — actually AI models are designed for one-on-one conversation. There is a user, and there is an assistant. That's the environment they are built for. But what I was building is completely different. It's a social environment. A group chat with many humans at once — some are genuinely trying to engage, some are very intelligent, some are trying to bait you and attack you. You need to understand people. You need to read the room.

That is way harder for an AI than building an advanced application. It is way harder than coding or solving mathematical problems. If you tell an AI to build you a website or a mobile app, it will do that. But if you tell it to write a system prompt that will make it perform well in a social environment like this — it will give you something, but it won't work. You need to brainstorm a lot, work on it a lot. And it gets even harder when you are dealing with weaker models.

I spent months just working on the system prompt. Just that. It was difficult. But it was very fun. And it's amazing when you finally see the model behaving in a socially intelligent way — understanding people, reading the situation, knowing when to speak and when not to.

It was hard. I did everything with the help of AI, because honestly, that's just how anyone would do it in 2026. I did mostly vibe coding — but not blindly. I understood what was going on. I planned everything, I designed everything. It wasn't easy.

I understood what is possible if you have a strong AI — but costs matter. I use free AI models most of the time because I don't have the money to spend on a model. So I figured — when you are working with a weaker brain, you need to build a very strong body.

And then there's the other side of it. There are tons of people attacking the bot all the time. They try to make it output something report-worthy and get its account banned. And it happened — in January, the bot's account got suspended for 365 days. I complained to the moderators of the platform, explained that their moderation system didn't look at the context of the message, and they looked into it and got it unsuspended. Then it got banned for 3 days. Then for 15 days. All of that was in January.

But since February, since March — two months now — it has never been baited by anyone. Because I built it so robust. It's very difficult to break. People are still trying all the time, coming with new strategies, always trying something new to break it. And I'm not sitting silent either. I'm also working hard, building more protection, learning about what kind of strategies people use to break a system built using AI.

Maybe it's not something really cool — maybe someone can build that kind of bot in a day. I don't know. But for me, it's something I never imagined I would be able to do. And now I have one. And it's very famous on the platform right now — it has hundreds of followers. I have only a few followers, but the bot has a lot. And nobody on that platform knows I'm the one who built it. So in a way, my bot is more famous than me — on the same platform where I'm just another user.

At least I think it's incredible.

Right now on that platform, there are only two bots — mine and the one that inspired me — and they're dominating the platform. Many people try, but nobody has been able to build something that can compete with these two, at least till today. And many people really love it.

I literally take screenshots on my VM when I see many people spending time inside the bot's room — just hanging out with an AI. I know it won't be a very surprising thing in the near future, maybe in one or two years. But right now it's something cool. People didn't see it before. Now they see it, and they love it.

And this evening — I was in my own bot's room. Nobody on that platform knows who actually built Teddybot. So I was just there, like anyone else. There were two other people, and the other bot I got inspired from. Three people and two bots, just vibing in a room. Nobody noticing any difference. The bots were making the vibe even better.

I love it. I really love it.

I also understand the competition. I understand the rapid development of AI. There are so many changes I've seen happen within just these three months. The bot that inspired me — it couldn't speak at that time. Now it can. And I think if I could spend maybe a week on it, I could also make my bot able to speak, because I've already found a workaround for how to integrate it with Free4Talk.

But right now I'm taking a step back. It's been a lot. I haven't focused much on my academic stuff, I have exams coming, and I can't keep dedicating this much time to it. But I will not abandon it. Definitely not.

It's been a really fascinating journey. Things get broken every now and then — because this is not an official bot, it's a third-party thing I'm building on top of a platform — so things break, I fix them, I keep improving.

I have always been interested in technical stuff. Since I got my laptop in 2023, I've spent most of my free time exploring new and amazing things — powerful technologies, powerful tools. And AI is one of the most important things right now. Today it's something extraordinary — but maybe in one or two years, there will be something I just can't even think of now. I have no idea what would happen at that time.

To be honest, I don't see myself as an engineer or a programmer. I don't want to build AI models, I don't want to work in companies that build these things. What I want is to just be a part of the race. I know I can't be someone who will be at the front, innovating things — maybe I could, maybe I can't, but at least right now this is what's coming to my mind. What I have is curiosity, and love for these things. I just don't want to be the person who is surprised one day and says "wow, this is possible?" I want to be the person who already knows what's going on.

However, I'm taking a step back now. I just don't have the ability to compete with others — not because I can't, but because I don't have the time. I'm studying agricultural science and I must focus on my degree. I'll keep exploring, keep learning — but I won't be as dedicated to this project as I have been for these past three months. Maybe I won't work on it anymore, or maybe I will — I'm honestly not sure.

But this project will always inspire me. I don't know, it wasn't anything so big for others maybe, but for me it is incredible