Code comprehension is the depth of an engineer's mental model of a codebase — not whether the code passes tests, but whether the person responsible for it can explain its behavior, trace its edge cases, and predict how a change will ripple through the system.
It has always mattered, but AI-assisted development raises the stakes. When an agent produces a working diff in seconds, it is easy to merge code that no human on the team actually understands. That "understanding debt" is invisible until an incident, a security review, or a refactor forces someone to reason about code they never truly read.
Measuring comprehension means checking understanding against evidence: can the author explain the reasoning behind an AI-generated change, anticipate its failure modes, and connect it to the concepts it relies on? Cartara turns each AI coding session into a short, evidence-linked comprehension check for exactly this reason.