When You Scale Output Faster Than Understanding
We assumed productivity and expertise were the same thing. We may have been wrong about which one we were building.
Someone once told me our generation, early-in-profession engineers, had it easy. We were lucky to have AI during undergrad as we prepared to enter the workforce. They weren’t wrong. AI lets us move and build faster than before. It explained concepts we hadn’t touched in weeks, debugged code, and filled gaps we had.
But we weren’t just using AI in a vacuum. We were competing against peers who were doing the same. That’s a completely different thing.
The bar didn’t stay fixed; it moved. What took a week became expected in a day. What was once exceptional became normal. Every productivity gain became everyone else’s gain too. The question was never, “Do you have AI?” It became: “Can you keep up when everyone does?” Suddenly, luck felt complicated.
Traditionally, productivity came after understanding. You struggled first, made mistakes, built intuition, and speed followed naturally. Strong foundations weren’t something you consciously optimized for. It simply came with the process.
We may be among the first generations experiencing that relationship in reverse. Now we hit walls, ask AI, find doors, and walk through them. We know enough to keep going, but not always enough to start over from scratch. Efficient and reasonable.
But some walls are worth sitting with, not because struggle is noble, but because certain insights only form after being stuck long enough to understand why things behave the way they do.
I learned web development a few years ago. I built my first calculators, small and ugly. I fought with centering divs. I spent hours debugging stupid mistakes. Those problems forced me to figure out how the pieces fit together.
Years later, when I had a 2000-lines application in front of me. While tackling bugs, I often knew where problems were likely to be, could open the exact file, and usually had a few guesses before reading the code. Over time, I had built enough mental models and pattern recognition to know what I was looking at.
Later, I built Flutter applications with AI for a course. I could produce working software, but I can’t honestly say I developed the same instincts. I didn’t spend enough time inside the underlying systems. I relied on AI to locate and fix bugs, which took many attempts and still didn’t leave me knowing why the problem had occurred. I could use the tools. I didn’t truly know them.
A 2024 study, How AI Impacts Skill Formation, found that AI assistance reduced performance and weakened knowledge acquisition in novice developers without meaningfully speeding their work, suggesting that output and learning do not always improve together.
Most of the time, the distinction between having access to an answer and knowing why it works doesn’t matter. Until the problem is unfamiliar, the tool isn’t there, or someone asks you to explain your reasoning from first principles. That’s when the difference between recognizing a solution and being able to reason about it becomes visible.
For experienced engineers, over-reliance on AI may just mean losing touch with the material. For someone still building those models, the tradeoff may be different.
Microsoft’s 2025 Work Trend Index, based on a survey of 31,000 workers across 31 countries, found that 53% of AI users struggle to verify AI outputs, and that heavy use was associated with declining confidence in independent judgment over time.
Many early-career engineers feel the gap between the speed they’re expected to operate at and the depth they actually have. They’re not blind to it. Some try to make time for deeper learning, an extra hour here, a longer postmortem there, but personal discipline can only go so far. You cannot fully opt out of the pace without, in some ways, opting out of opportunity.
This isn’t a wholesale indictment of AI. The tools have given unprecedented capabilities and opened new kinds of learning. The question is narrower but, I think, important:
What kind of professionals are we building when productivity routinely arrives before understanding?
It’s a practical worry about the foundations we lay beneath our work. Today’s early-career engineers will eventually become the people maintaining critical systems, leading teams, and making decisions under uncertainty. If those foundations are thinner than we realize, we’ll discover it when we can least afford to.