AI is shifting engineering value from execution to problem selection

When chess engines surpassed human grandmasters, something unexpected happened: for a period of time, the strongest players weren’t humans or machines, but combinations of both — “centaurs.”

The engine handled calculation. The human decided which lines were worth calculating.

As engines improved, they absorbed that layer too, and the human role moved up the stack.

I think something similar is happening in engineering.

AI tools are making it much cheaper to produce code, documentation, and designs. As a result, output quality is becoming a weaker signal of experience than it used to be — not because expertise matters less, but because it shows up differently.

The constraint is shifting toward deciding which problems are worth solving, spotting flaws before implementation, and asking the right questions early.

When producing outputs becomes cheap, the value moves toward deciding what should be produced.

This doesn’t mean junior and senior engineers are equivalent — system-level thinking still separates them. But the surface signals organizations use to detect that difference are compressing quickly.

Curious if others are seeing this in how they hire and evaluate engineers.