The Senior Developer's Case

A 10-year software engineer posted about building side projects entirely with AI. From a phone. Without reading a single line of the generated code.

Their argument was not that code review is unnecessary. It was that experience gives you a mental model of correct that does not require reading every line. "I know what good code smells like. I know the architecture I would build myself. So I know when the AI has gone off-track , not from reading the code, but from running it and watching where it fails."

This is defensible. For a 10-year engineer working on a side project with no production users, no compliance requirements, and no team that will inherit the codebase. For that specific set of conditions, the risk profile of vibe coding is acceptable.

The problem is how the conclusion gets borrowed from those conditions and applied everywhere else.

What the Speed Actually Costs

The judgment that the 10-year engineer is relying on when they "know when the AI has gone off-track" was built during the years when they read every line of code. The catches they make on feel , the architecture that looks right, the bug that smells wrong , are the accumulated product of years of close reading.

You cannot shortcut to that judgment. You build it by doing the slow thing first. By reading the code you did not write. By understanding why something breaks, not just that it broke.

The senior engineer giving up code review is spending saved capital. The junior developer starting with vibe coding never builds the capital to spend.

The Addiction Pattern

Speed is genuinely addictive. Shipping features in an hour instead of a day is real. The feeling of continuous progress , always moving, always producing , is real. Comparing this to the slowness of careful work makes the careful work feel like a productivity problem rather than a discipline.

The vibe coding addiction is the belief that the speed is the point. The speed is a side effect. The point is shipping something that works, that you understand, that you can fix when it breaks at 11 PM. The speed that comes from not understanding what you shipped is not speed. It is debt.

When to Vibe Code

Prototypes that will not reach production. Side projects with no users who depend on them. Throwaway experiments. Anything where "it works right now" is sufficient and "it will work in six months when someone else has to debug it" is not required.

That is a real category. It is not nothing.

It is also not most of what gets built in production.