There's No Limit to How Bad Code Can Get

There's a persistent metaphor for neglected codebases: a sinking ship. It implies an eventual end — a point where the water hits the deck and everyone abandons. But that's misleading. As Zach Kehs argues in his essay There's No Limit to How Bad Code Can Get, unlike a physical ship, a software system has no inherent structural limit. You can keep adding complexity indefinitely; it won't collapse. It'll just get slower, harder to change, and more painful to work in.

A real-world example from Amazon

Kehs describes his decade-old experience on an order-processing team at Amazon. The system's core function—writing to a database and calling other services—should have been maintainable by two dozen engineers. Yet the org had hundreds of people. Institutional knowledge eroded as engineers left, leaving 'haunted graveyards' in the code. Business rules were often undocumented, and the system spanned team boundaries, making tracing impossible.

Every few years, a new senior hire would attempt a re-architecture. It always failed. The system was too complex to understand fully, and the impatient promotion cycle meant fixes were designed with incomplete information. The result: new layers grafted onto the architecture, headcount increases that never left, and the cycle repeating.

Why the metaphor breaks down

The author's key insight: a business can sink — but the software itself won't. The code can always get worse. There's always another layer of indirection, another performance degradation. But ironically, the business often dies before the code reaches any kind of 'bottom.'

For developers dealing with legacy systems, the takeaway is that 'sinking ship' gives false comfort. There's no point of no return where things magically reset. If you're in such a system, you're not waiting for it to sink — you're waiting for someone to decide the cost of change outweighs the cost of maintaining the status quo.

Read the full essay for a deeper reflection on the mechanics of technical debt in large organizations.