Blog

Instinct Over Experience: How Great Devs Work With AI

Back to blog

It’s Not About Knowing Everything. It’s About Knowing When Something’s Wrong.

There is a persistent myth in engineering that the best developers are the ones who know everything. Every file. Every edge case. Every undocumented behavior buried somewhere in production. That model breaks down fast in an AI-assisted world.

The next generation of strong developers is defined by something else entirely. They are good at noticing when the AI is off. Not wrong in an obvious way, but subtly wrong. Answers that technically work but feel fragile. Fixes that pass tests but create risk. Confident responses that do not line up with how the system actually behaves.

That instinct matters more than encyclopedic knowledge. You can teach syntax. You can document systems. You cannot easily teach the feeling that something does not smell right.

You Are Not Good Because You Know the App

Knowing the app helps. It is not the differentiator anymore.

What matters is understanding what normal looks like. How data usually flows. Where latency shows up. What logs look like when things are healthy. When an answer fits the shape of the system and when it does not.

AI can read the repo. It can summarize the architecture. It can propose fixes faster than any human. What it cannot reliably do is judge whether a solution fits the real world constraints of your system. That judgment still belongs to you.

That is the skill gap now. Not knowledge. Discernment.

The Pilot Model

The way we think about this internally is simple. We are pilots.

We handle takeoff and landing. Architecture. Risk. Decisions that are hard to reverse. We set direction and define boundaries.

The copilot handles most of the middle. Routine execution. Investigation. Mechanical steps. The long stretches where things are predictable.

Until they are not.

When something goes sideways, instinct takes over. You know when to disengage. You know when to dig deeper. You know when a clean looking answer is hiding a bigger problem.

That is not experience measured in years. That is experience measured in attention.

Why Role Based Automations Matter

We use automations heavily, but not in a naive way. The goal is not to let AI do everything. The goal is to make sure it does the right things in the right context.

Different roles require different behavior. A developer should be able to investigate and fix issues. A tester should validate and report without touching code. A product mindset should analyze scope and tradeoffs without implementing anything.

The important part is not the tooling. It is the constraints.

Each role has clear permissions and clear stop points. That is what keeps AI useful instead of dangerous. It reduces accidental overreach and makes behavior predictable.

Predictability builds trust.

The Unexpected Benefit

This setup does not make people lazier. It does the opposite.

When AI handles the mechanical work, humans are pushed toward judgment. Toward asking whether something actually makes sense. Toward understanding impact instead of just correctness.

You stop rewarding people for memorizing systems. You start rewarding them for taste, for restraint, and for knowing when to slow down.

That is what scales in a world where execution is cheap.

The Real Point

You are not valuable because you know everything.

You are valuable because you know when something feels wrong, even if you cannot immediately explain why.

Models will keep improving. Tools will keep getting better. The ability to sense risk and step in at the right moment will remain rare.

That instinct is the job now.

Have a project in mind?

Lopie is a senior-led software studio in San Diego that builds custom software and AI systems. If you have a project in mind, tell us about it.