AI has become an assumed part of almost every business pipeline. It’s no longer a question of whether to integrate it into a workflow, but how - and done well, it can meaningfully improve the speed and quality of what a team delivers.
But last night I was talking with Brandon Kissinger, and one analogy has stuck with me since. He compared AI to a high-quality pair of scissors handed to a barber. In the hands of someone inexperienced, those scissors produce a catastrophic haircut. In the hands of someone skilled, they become a genuine productivity multiplier. The tool doesn’t change but the outcome depends entirely on who’s holding it.
Too often, though, AI isn’t treated as a tool at all. It’s treated like a genie: state your wish, and it fulfills it. Watching that go wrong is almost funny, until it’s your production system. Most engineering work is complex and full of trade-offs. Engineers need to be able to make decisions based on those trade-offs and anticipate how an implementation will actually behave once it lands in production with no surprises.
That decision-making skill comes from experience. Real, hands-on experience, backed by knowledge and by deliberate time spent in the lab, experimenting with concepts and ideas rather than just shipping. This is what I see most people skip when they lean on “vibed” applications. Foundation matters, and it shouldn’t be dismissed.
The appeal of vibe coding is real, especially when you treat the AI as a genie somehow it can feel like doomscrolling. But our brains aren’t wired to work like machines. They’re slow. They need time in the zone to do real thinking. AI, by contrast, is fast: open the sidebar, and it’s already working. Our brains love a shortcut, and they’ll happily hand the wheel to genie mode — which blocks the deep cognitive reasoning you need to actually understand your system.
To stay in control of a project, control over how you use AI is essential. Going back to the analogy: the scissors don’t decide when to cut. The barber’s hand decides when to apply pressure. Here’s how to keep your hand on the scissors.
1. Take ownership
Owning a task, feature, or product changes how you relate to the work. When something you’ve poured effort into performs the way you hoped, it pays off in a way that outsourcing it never quite does the same way a meal you cooked yourself feels different from one you bought, even if the food tastes the same. Ownership also breeds curiosity: once something is yours, you want to know how to make it better, not just how to make it work.
2. Treat understanding as the core process
The brain isn’t a solid-state drive, it can’t just absorb and store information on contact. It takes real effort to map new knowledge into what you already know. When you hit a problem, it’s often faster to just prompt your way past it, but try holding off on that instinct. Open the docs. Read the post that explains the concept you’re actually stuck on. Approach it with an empty glass: you’ll usually find more context than you strictly needed for the immediate problem, but that surrounding context, the edge cases, the “why,” the adjacent use cases is exactly what turns a one-off fix into real understanding.
3. Practice deliberately, not just reactively
Don’t let every interaction with a new concept be triggered by a production bug or a deadline. Set aside small, low-stakes sessions to experiment, build a toy version of the thing, break it on purpose, read the source of a library you depend on. This is the “lab time” that turns borrowed knowledge into your own intuition, so that when a real trade-off shows up under pressure, you’re recognizing a pattern instead of learning it for the first time.
4. Review before you accept
Before merging AI-generated work, read it the way you’d review a teammate’s PR: ask why it made the choices it made, not just whether it runs. If you can’t explain a piece of code you’re about to ship, that’s the signal to slow down, not speed past. This single habit is often the difference between using AI as a tool and quietly handing it the scissors.
None of this is an argument against using AI, it’s an argument for staying the one who decides when to cut. The scissors are genuinely good. But a good tool in inexperienced hands still produces a bad haircut, and the only way to build that experience is the slow way: through ownership, understanding, and deliberate practice. Skip the foundation, and you’re not doing engineering anymore, you’re just hoping the genie is in a good mood.