Ein Bier, gebraut mit KI: Wie zwei Österreicher mit Lobster Lager im Silicon Valley landen

·33 min·2 clips

A Saturday night call from NVIDIA's Swiss team suddenly put Lobster Lager in the running for Jensen Huang's keynote.

1. 'Was wichtig ist,' the Die Presse news podcast, devotes a Friday episode to Lobster Lager — the story of how an AI-agent beer project by two Austrians reached the stage of NVIDIA's GTC keynote in San Jose. 2. Host Anna Wallner interviews Stefan Erschwendner, a Salzburg-born management consultant who has lived and worked in Berlin for 14 years, and whose father Gerhard lives in the Swiss village of Gelterkinden in the canton of Basel. 3. Lobster Lager is framed as an experiment: can an AI agent (OpenClaw) run the entire process of a microbrewery — from brainstorming ingredient combinations to ordering supplies, controlling the brew tank, building a website, running an online shop, and managing newsletter and social media — with minimal human input? 4. The name 'Lobster' comes from the lobster emoji and icon used by the OpenClaw community, created by OpenClaw developer Peter Steinberger and embraced by the project. 5. Stefan's 68-year-old retired father — described as having taken up home brewing as a retirement hobby — did approximately 95% of the technical work on his own, including writing the BrewOS programme and independently adding a webcam to give the AI agent visual monitoring of the brew tank. 6. Stefan's company helped configure OpenClaw at the start and provided occasional support; the father worked autonomously and would inform them after completing each development step. 7. The AI agent places ingredient orders, monitors temperature via Bluetooth sensors in the brew tank, oversees the brewing process visually via webcam, and builds website, online shop, and newsletter infrastructure — with the human's only required task being physical actions like adding hops and connecting water lines. 8. The project originated as a response to a call for use cases at ClawCon Vienna, a gathering of the OpenClaw fan community at the Prater, which coincidentally occurred one week before it became public that Peter Steinberger would join OpenAI. 9. Stefan emailed Steinberger with a brief description of Lobster Lager and received the reply 'I love it, let's do that' — Steinberger then brought the project to Silicon Valley without Stefan knowing. 10. On a Saturday evening, Stefan was at dinner when his father called saying NVIDIA Switzerland had reached out; Stefan was transferred through successive NVIDIA teams — Swiss, then San Francisco — until at approximately 9pm he learned the project was being considered for Jensen Huang's GTC keynote. 11. Madison Huang, Jensen Huang's daughter and part of the NVIDIA team, specifically requested that someone from Lobster Lager be present in San Jose; Stefan booked a transatlantic flight at 3am, slept briefly, and flew out at 10am. 12. In Jensen Huang's approximately 2 hour 20 minute GTC keynote, Peter Steinberger described a 60-year-old man who connected his brewing machine to OpenClaw via Bluetooth and automated everything including the website — approximately 30 seconds of the global presentation. 13. ChatGPT, when later asked, estimated the probability of a project being featured in Jensen Huang's keynote as lower than winning the lottery. 14. Post-keynote, Stefan's consulting firm has signed its first American client in Louisiana, received Venture Capital inquiries, attracted interest from breweries and food conglomerates, and is in discussions with NVIDIA itself about brewing a batch for their headquarters as a customer gift or staff beverage. 15. The episode positions the Lobster Lager story within a wider cluster of Austrian AI contributions: Mario Zechner's PI coding agent (described as the foundational layer OpenClaw is built on), Peter Steinberger's OpenClaw itself, and Andreas Klinger's EU Inc. framework advocacy. 16. Anna Wallner references tech journalist Elisabeth Oberndorfer's WZ column lamenting that while Jensen Huang has called OpenClaw 'the next ChatGPT,' European companies are not building on it — Silicon Valley firms are. 17. Stefan's philosophical position on AI risk: the biggest danger is not trying. He argues people need to experiment incrementally, starting with low-stakes projects, to develop intuition for what agents can do. 18. The episode is conducted in German, with four embedded English-language audio clips: Peter Steinberger at GTC, JD Vance-era clips, Kevin Roberts, and Scott Hanselman's podcast interview with Steinberger. 19. Listeners interested in AI agents, OpenClaw, the practical use of LLM-driven automation in physical processes, or the Austrian tech ecosystem will find this episode directly useful. 20. Listeners expecting rigorous risk analysis or critical examination of AI safety will find the episode optimistic and light on sceptical counterarguments.

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Was wichtig ist