CZPodcast

CZPodcast is a Czech-language technology show built around fast conversation, working examples, and guests who know the machinery from the inside. It has the feel of an old developer room with microphones added. The regular voices joke about episode numbers, recording gaps, events, merch, and their own process, then pivot into serious questions about AI products and software infrastructure. That mix is the texture of the show. Recent episodes are heavily focused on applied AI. Microsoft stories turn into a discussion of automated tools, agents, internal adoption, sales organizations, and how much trust companies are being asked to place in automated tools. A conversation with Tomáš Srnka frames E2B as sandbox infrastructure for agents, Linux workloads, reproducible environments, compliance-sensitive processing, and generated artifacts. Another episode looks at AI backend generators, contrasting component generation with data models, PostgreSQL, Node.js, GraphQL, React interfaces, and the boring work teams want to remove from client meetings. Petr Baudyš brings the discussion into Rossum.ai, document processing, transformers, few-shot learning, user corrections, invoices, and the decision to push a company into a new technical direction. Matouš Havlena adds a sharper look at agent systems, orchestration, Cloud Code, repositories as company memory, MCP, skills, workflows, and the security risks that arrive when agents become more capable. The show is not tidy radio. Questions stack up. Answers restart. Someone interrupts to clarify a term, make a joke, or drag an abstract claim back to a concrete deployment problem. That informality works because the technical center is strong. CZPodcast is especially good at exposing the gap between an impressive AI demo and something a company can run, trust, sell, or maintain. It treats hype as material for questioning, not as a conclusion. The result is a conversational map of Czech software culture at the moment when AI is moving from novelty into infrastructure.