AI is accelerating how quickly we can build software, but the long-term consequences of our decisions haven't disappeared. As implementation becomes easier, sound technical judgment becomes even more valuable. Leaders need to ensure engineers develop that judgment through exposure to production realities, not just new development.

Most of the signals that shape an engineering career, from interviews to design reviews to promotion packets, celebrate designing and launching new systems more visibly than maintaining and evolving existing ones. As a result, engineers often move on to the next project before they can learn from the long-term consequences of their decisions, and they may underestimate the operational and organizational costs their next design will create.

Those who do take on legacy work face their own challenges: understanding accumulated dependencies and operational burdens, identifying high-leverage improvements, and making the value of that work visible across the organization.

I’d encourage attendees to pick one system they maintain or have inherited and work through three questions. What assumptions have proven wrong in production over time? What operational burdens have accumulated around it, and who is carrying them? And what would you design differently if you were starting today? The answers often point to the highest-leverage improvement to make next, and they become the judgment you can carry into the next system you design.

The Technical Judgment Field Guide I’ll share at the talk includes worksheets for going deeper.

I appreciate QCon's emphasis on practitioner-led talks that go beyond technical implementation to explore real-world engineering decisions, tradeoffs, and lessons learned. Its focus on experienced engineers and technical leaders creates an opportunity for conversations that connect architectural choices with the organizational realities of building and operating software.

I still remember Jessica Tai’s talk at QCon San Francisco 2018, “Airbnb’s Great Migration: From Monolith to Service-Oriented.” At the time, moving from monoliths to services was the hot architectural trend, much as AI is today. Looking back, it’s interesting to see how the technologies that capture our attention evolve over time, while the fundamental challenges of making sound engineering decisions remain remarkably consistent.