AI-generated code suffers from a hidden visibility inversion where the deepest, most dangerous errors are also the hardest to find.
AI-written software undergoes a unique form of decay because the model does not understand the long-term state of a system. This structural degradation happens because AI focuses on local probability rather than the overall health of the codebase. The result is a system where small, superficial bugs are rare, but catastrophic architectural flaws are baked into the foundation. This theory moves beyond simple hallucinations to explain how AI can fundamentally break the architecture of large software projects over time. Developers must now watch for silent decay that traditional testing tools might miss entirely.
Structural Degradation: From Stateless Generation to Layered Software Decay in AI-Generated Code
SSRN · 6655438
Generative AI introduces a fundamental structural mismatch into software engineering: code is produced through stateless probabilistic generation, yet integrated into systems that evolve through stateful, temporally constrained architectures. While existing discourse concentrates on hallucinations, isolated defects, and productivity gains, it lacks a unified account of the degradation mechanisms unique to AI-assisted development-mechanisms that are neither reducible to individual bugs nor adequa