Member-only story
The Vibe Coding Reckoning: What Happens After the Code Works?
Getting software to work is becoming dramatically cheaper. Understanding what happens when it fails is not.
Not a Medium member? You can read it for free using the official Friend Link.
A developer needs an internal customer-support dashboard. They describe what they want to an agent, and by the end of the afternoon the application has authentication, customer search, filtering, database access, CSV export, a polished interface, and deployment configuration.
The developer clicks through the main flows. Everything seems fine. Two visual issues appear, the agent fixes them, the tests pass, and the application ships.
By almost any reasonable standard, this is a success.
Then six weeks pass.
A customer record appears twice after an import. Someone asks whether rerunning the job is safe. Another engineer notices that two different fields seem to represent the customer’s current status. Deleting an account removes the visible profile but leaves related records behind. During an incident, a background task times out and nobody is immediately sure whether retrying it will repeat an external side effect.
None of this means the original application was fake or that AI somehow failed to build…
