I’ve just kicked off another round—the fourth edition—of AI_Devs (since March 9, 2026). 🚀 I’ve been in it from the very start (since 2023), which gives me a unique perspective on how fast work with large language models (LLMs) has been evolving.
Evolution of paradigms 🔄
The AI_Devs initiative, led by Adam Gospodarczyk, Jakub Mrugalski, and Mateusz Chrobok, is now the key reference point in the Polish IT ecosystem for anyone who seriously wants to combine generative AI with application logic.
Looking at the four stages of this cycle, you can clearly see how far the IT industry has come:
- AI_Devs 1 (2023) – the pioneering edition, focused mainly on prompt engineering and simple text-generation scripts. 🪄
- AI_Devs 2 (Reloaded) and AI_Devs 3 (Agents) – the shift from calling APIs to building logic and the first agentic interactions. 🤝
- AI_Devs 4 (Builders) – the latest edition, where we step into advanced, autonomous agent ecosystems operating within complex communication protocols, such as the Model Context Protocol (MCP). 🧠
This is no longer just “chatting with a bot.” It’s about building systems that analyze, plan, and execute complex tasks within our applications.
From learning to product: Charopis 🛠️
For me, returning to AI_Devs is also a bit of a sentimental journey. It was during the first edition, while learning the basics of working with LLMs, that I started laying the groundwork for Charopis.
What began as a simple experiment and manual prompt testing in Postman grew into a dedicated product engine. Today, Charopis is deployed in production at Skład and provides huge independence. It lets us generate consistent product descriptions at scale, build our own individual style of communication, and ultimately strengthen the store’s authority with customers. We turned the copywriting bottleneck into a predictable, automated process.
I see my participation in the fourth AI_Devs edition as fuel for further growth. Autonomous agents and standards like MCP open entirely new integration possibilities, which will certainly show up in my next implementations and in the ongoing evolution of Charopis.
Back to building!