Why Agentic AI Reads Legacy Code the Way an Archaeologist Reads Ruins, Not the Way a Compiler Does

A compiler tells you what code executes. An archaeologist tells you what the people who built it were trying to accomplish. For most enterprise systems, only one of those is actually useful. Picture this. A senior engineer gets pulled into a meeting about upgrading a core transaction system. Halfway through the proposal, they say something every CTO has heard at least once: "We can't touch that module. Nobody knows what it does." The system still runs. The logic still executes. But the reasoning behind it — the regulatory interpretation, the deliberate design choice made by an architect who left eight years ago — is nowhere to be found. It lives only in the behavior of the code itself, waiting to be misunderstood or quietly destroyed by the next developer who assumes they understand it. This is not a maintenance problem. It is a comprehension problem. And the tools most organizations use to address it were never designed for it. What a Compiler Sees — and What It Misses Traditional cod