Matthew Hansen published a sharp piece on what AI tools actually change about software development, and the argument applies well beyond coding.
The core thesis: writing code was always the easy part. The hard part is reading it, understanding it, reviewing it, and knowing whether it’s correct in context. When you hand the easy part to AI, you’re left with nothing but the hard parts — and you’ve lost the understanding you would have built by doing the work yourself.
The Problem
Hansen describes a pattern anyone using AI tools has seen:
- An AI agent confidently deletes 400 lines from a file while claiming it changed nothing.
- “Vibe coding” works great for prototypes and falls apart under real constraints.
- Developers ship code they don’t fully understand because “AI did it for me.”
None of this is hypothetical. It’s the same risk as copying from Stack Overflow without reading the answer — except now it happens faster and at greater scale.
The Productivity Trap
One fast delivery resets management expectations. The next sprint gets scoped larger. Developers skip tests to keep pace. Technical debt accumulates. The initial productivity gain disappears into a cycle of escalating commitments and declining quality.
Hansen frames it well: AI has senior-level skill but junior-level trust. It can produce competent code, but it has no organizational context, no institutional knowledge, and no accountability. That review burden falls entirely on the human.
Where AI Actually Helps
The article isn’t anti-AI. Hansen describes using AI effectively during a production incident — a timezone bug across systems — where he provided the context and the AI handled research and investigation. That’s the right model: use AI to tackle genuinely hard problems where you can verify the output, not to skip the learning that makes verification possible.
Our Take
We see parallels in IT operations. Automation and AI tools are powerful, but they don’t replace the need to understand what’s happening in your environment. A monitoring dashboard you don’t understand is just a screen with lights on it. The value comes from knowing what the alerts mean and what to do about them — and that knowledge comes from doing the work.