This session shows why software engineering expertise is currently the key to turning AI agents into real, production-ready systems.

A new opportunity has emerged specifically for experienced developers: the ability to design, build, and evolve real, production-ready applications using AI agents - not just to speed up coding, but to create complete software systems for personal productivity. This capability is still largely underutilized and requires a developer’s understanding of architecture, constraints, and system design to guide agents effectively.

In this session, I will show how I use Google Antigravity, an agent-based AI product by Google/DeepMind, to build and operate real Progressive Web Apps using a familiar stack: React, Tailwind, Firebase Hosting, and Firestore.

One example is a custom PWA for managing my triathlon training. Instead of focusing more on training, I did what many developers instinctively do - I built a better tool tailored exactly to my needs. I applied the same approach to travel planning. My wife and I frequently plan weekend trips involving flights and accommodations, so I built a second application - using the same stack - to create a unified, structured travel overview.

These are not prototypes or scripted demos. You will see how AI agents are instructed with developer-level intent, how architecture emerges through real usage, and where classical software engineering remains essential to keep systems maintainable and robust.

The session concludes with a live, on-stage evolution of one application: adding a real requirement, adapting the data model, and improving the logic - openly and without shortcuts.

This talk is not about replacing developers. It shows why this moment is especially powerful for developers, and how they can leverage AI agents today to build production-ready personal software systems that go far beyond simple automation.