With all the recent advancements in AI, we found ourselves, almost inevitably, uncovering better ways of developing software by doing it and helping others do it. This led us to a fundamental question: Can a pair of Product Managers and Designers build something real, reliable, and production-ready with minimal developer supervision? We wanted to move past the marketing hype and find out if non-engineers could actually navigate the complexities of a live codebase without a developer holding their hand at every step.

To find the answer, we ran a dual-track experiment following two PM/PD pairs on very different journeys: one starting a project from scratch and another attempting to build and ship products within a massive, pre-existing production environment filled with established patterns and strict architectural rules.

In this talk, we will pull back the curtain on the entire process. We will explain how we configured the initial setup to enable autonomy, the specific technical hurdles and logic-based problems we encountered when the AI hit its limits, and the crucial moments where the experiment failed and developers had to intervene to save the day. Finally, we will share how we assessed our confidence in the outcome—whether we truly reached a level of trust where the final code could be deployed—providing an honest look at whether the "Product Engineer" is a sustainable reality or a dangerous myth.