Who I am
My name is Vojtěch Pszczólka. I'm based in Prague and IT is my main field. I have experience as the IT manager of a mid-sized company and as a systems engineer on critical infrastructure. The arrival of AI, though, activated my ADHD superpowers.
I got seriously into AI on July 18, 2024, the day gpt-4o-mini came out: it was smart enough and at the same time cheap enough that I could start using it via the API at scale. I spent my first tokens - and ever since I work with AI every day, constantly finding out what else it can be used for and how. I use it for everything, and only afterwards examine how much sense it makes. In 2025 alone that meant over 30 projects, thanks to vibecoding. Eventually I handed part of the infrastructure over to the agents; they now manage the whole path from the firewall to the very code running on this web server.
I enjoy explaining complicated things simply. I like asking myself hard questions. And I can say "I don't know" - which I appreciate in others too.
Outside work I enjoy sci-fi, physics, electronics, and any system where something connects to something else. Yes, including computer networks. Especially computer networks.
Values
I consider myself a libertarian - maybe even an anarcho-capitalist - an existentialist and an individualist. My central value, though, isn't freedom; it's intellectual integrity. I need it to be able to answer my own questions properly. Without integrity a person has no honest information to work with, and someone deciding on manipulated information isn't deciding freely.
I'm an active member of Libertyloft.cz, where I ran the experiential event Konsensus.
What I do
- Business IT: Microsoft 365, SharePoint, Power BI, ERP/CRM, users, and security.
- Practical AI: chatbots, automation, wiring language models into company systems.
- Skills: networking, Linux, storage (SAN), virtualization, backups; failover, load balancing, clustering.
- Games and live experiences: detective games, LARPs, and teambuilding events where the player lives out their own story.
- Workshops on the practical use of language models - in companies and in negotiation.
Projects
- chatmystery.com - a text-based detective game in English. You don't click on clues - you interrogate AI characters, each with its own memory, motivations, and version of the truth. My very first project built on an LLM API. It became obsolete before we managed to finish it.
- Hra na vraždu (The Murder Game) - a detective LARP for companies and groups. Participants don't get a task; they get a role, a motivation, and a situation they have to handle with the others. The result is an experience and the release that comes with improvising.
- mujkviz.cz - a platform for independent pub-quiz organizers: questions, presentations, answer sheets, promotion. The modules go by names like Pan Myšák, Vševěd, or Sokol, and the AI question generator already has real users.
- RoboVojtík - an internal AI Teams chatbot for company IT support. I also used it for demos and for training on what AI can help with.
- iVojtík - an attempt at an autonomous agent that was supposed to improve itself entirely on its own. No one in the world really has that technology yet, but it taught me a lot.
- sinuhetcloud - personal infrastructure: two cloud ARM nodes named Romeo and Julie, a Proxmox node called Cain, the Vtěrka firewall, the Beruška Raspberry Pi, the agents Bořek, Edgar, and Dejv, and the Hromádka database. I enjoy naming things.
- LARPs and scenarios - a medieval game about the budget of Charles IV, Konsensus, an educational game about political manipulation, and Pszcz! - a game about a secret society defending digraphs against Jan Hus. With a surname like mine, it was a duty.
- AI manifesto - my dystopian vision of AI, captured in three songs. A YouTube playlist.
Talks
In 2025 my team and I took on an AI negotiation challenge under the Harvard banner - I built its multi-agent architecture (that went well) and negotiated in person too (that went less well).
A few months later I co-led a workshop at the European Negotiation Conference at the Sorbonne in Paris, on using language models in negotiation training.
Contact
- E-mail: vojtech@pszczolka.cz
- GitHub: github.com/procmadatelzobak
- LinkedIn: Vojtěch Pszczólka