Post by Sebastian Baltes

Thanks again to everyone who provided input for our group retreat. It is, of course, hard to condense all discussions into one post, but I want to at least share the themes we developed to structure our conversations: 𝐎𝐫𝐠𝐚𝐧𝐢𝐳𝐚𝐭𝐢𝐨𝐧 𝐚𝐧𝐝 𝐩𝐫𝐨𝐜𝐞𝐬𝐬𝐞𝐬: How does the team and organizational setup change? How do we handle the asymmetry between software that is cheap to create but expensive to understand, verify, and maintain? 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠 𝐚𝐧𝐝 𝐜𝐫𝐚𝐟𝐭: How do we integrate AI into software engineering workflows to reliably create better software? Which deterministic checks can help control increasingly non-deterministic workflows? When do we need additional checks, and when can an AI agent work independently? Which context improves outcomes, and which context does not? Will software development be reduced to working with natural language text? 𝐂𝐨𝐠𝐧𝐢𝐭𝐢𝐨𝐧 𝐚𝐧𝐝 𝐬𝐤𝐢𝐥𝐥𝐬: How does the role of junior and senior developers change? How does ownership change, beyond the legal sense? When should we pause? What are the long-term effects of cognitive offloading? How do developers decide what to believe, and what not to believe, about the capabilities of AI tools? 𝐑𝐞𝐮𝐬𝐞: Does code quality still matter? How do we handle and verify the one-off tools that AI agents write and run during sessions? Does agentic AI lead to a new category of software projects: “throwaway software”? Are code clones acceptable now? So much we still don't know, but at the same time so many opportunities for research. :-) Brenda Ayuku Chiteri Moein Mohsenimofidi Alexander Berndt Tobias Reichel Timo Speith Shalini Chakraborty Vahram Nersesyan Klaus Häuptle Daniel Dahlmeier Hannes Rüskamp Daniel Gnad Dr. Oliver Kopp Greg Wilson Zhiyang Chen Andreas K. Rainer Koschke Marlon Kranz Fabian Peña