Comprehension debt: the real risk in buying code nobody understands
Vibes Crow3 min read
You've heard of technical debt: shortcuts in a codebase that make future changes slower and riskier. AI-built software introduces a sharper, less-discussed cousin — comprehension debt. It's not that the code is messy. It's that no human understands how it works — often including the person selling it to you.
This is the defining risk of acquiring a vibe-coded app, and almost nobody prices it.
Why AI-built code is uniquely prone to it
When a developer writes code by hand, understanding accrues as a byproduct. They had to reason through every decision, so they carry a mental model of the system. That model is most of what you're really buying when you acquire software.
AI-generated code skips that step. The app can work beautifully while the "author" only ever reviewed the output at a glance. The mental model was never built — not by the AI (it doesn't retain one), and frequently not by the seller either. You're buying a working machine with no operating manual and no one who can draw the schematic.
What comprehension debt costs you
A working app with no one who understands it is fine — right up until you need to change it. And you will:
- A dependency hits a breaking change and the build stops.
- A payment edge case starts silently failing and you can't find where money is handled.
- You want to ship the obvious feature that doubles revenue, and you discover that touching one part breaks three others for reasons no one can explain.
Each of these is routine for an owner who understands their system. For an owner who inherited a black box, each one is an existential event. The asset's value isn't the code — it's your ability to operate and evolve the code. Comprehension debt is a direct discount on that.
How to detect it before you buy
You can't grep for "nobody understands this," but you can find its fingerprints:
- Ask the seller to explain a specific flow. Not the pitch — the mechanics. "Walk me through what happens when a user's card is declined." Fuzziness here is the single loudest signal.
- Look for tests. Tests are codified understanding. Their presence means someone reasoned about correctness; their absence means the system's behavior is whatever it happens to do today.
- Check the bus factor. One person who half-understands the system is a bus factor of roughly zero. Map how many humans could actually keep it running.
- Trace IP and provenance. Who or what generated this, and is there any record? A codebase with no history is a codebase with no memory.
These map directly to the checks in our Asset Report — code health (do tests exist?), bus factor (how many real points of failure?), and IP provenance (is there a trail?). They're in there precisely because comprehension debt is the risk that screenshots and revenue charts can't show you.
The reframe
Comprehension debt isn't a reason not to buy AI-built apps — it's a reason to buy them with eyes open. A black box at the right (lower) price, with the risk acknowledged and a plan to build understanding, can be a great deal. A black box sold as a turnkey business at a turnkey price is how acquisitions turn into write-offs.
The whole point of verifying an app before it changes hands is to convert "trust me, it works" into "here's exactly what you're inheriting" — comprehension debt included.