Software Engineering Beyond the AI Hype
“I’m not changing jobs; I’m just participating in a longitudinal study of faculty onboarding.”*
Before people start getting nervous: No, I’m not intending to move again 🙃
However, my inaugural lecture in Heidelberg is around the corner. If you are available on Wednesday, 10th June, between 13:15 and 14:00, you are invited to join me in the lecture hall of the Mathematikon building in Heidelberg. Update: The slides and the recording of the lecture are now available.
Abstract: Headlines proclaim that “coding is solved,” and “vibe coding” suggests we may soon forget that source code even exists. Meanwhile, even self-described “vibe coders” at major tech firms still rely on design review, code review, and staged deployment. Deleted production databases and leaked customer data are the result of skipping those practices and pushing straight to production. This talk argues for an evidence-based view of AI for software engineering, organized around four claims: (1) software engineering is far more than programming; (2) “solving” programming does not solve software engineering; (3) today’s agentic AI assistants are a software systems innovation more than a model innovation; and (4) the debate needs far more empirical evidence, both to back its claims and to engage the evidence we already have. Drawing on my experience across industry and academia, I make the case for empirical research that studies both the developers who use these tools and the tools themselves, to rigorously assess their benefits and limitations.
* I asked ChatGPT to come up with jokes about my recent moves (SAP → University of Bayreuth → Heidelberg University), and this is what came out.