Generating code quickly is not the same as building software that lasts. The real risk is not output. It’s liability.
Cory Doctorow’s recent article,
Code is a liability (not an asset), makes a case worth sitting with. Code is not an asset simply because it can be produced at speed. It only creates value if it keeps working as requirements shift, infrastructure changes and edge cases emerge. Every system, over time, accumulates dependencies and workarounds. Good software does not just work at launch; it adapts. More importantly, it fails gracefully when the unexpected happens.
The concern with AI-generated code is not the volume. It’s the long tail. Faster output brings with it more maintenance, more refactoring and more governance overhead. If the only metric being optimised is speed, the problems do not disappear. They compound. As Doctorow puts it: AI code written at 10,000 times human speed, designed to work but not to fail gracefully, is the digital equivalent of filling your walls with asbestos. Future teams will spend years removing what we built in minutes.
There is a difference between writing code and engineering software. Confusing the two is not just a technical mistake. It is a business one.
