If coding is solved, what now?: Measuring the sloppiness of code

LLMs can generate syntactically correct code, but they often introduce unnecessary abstractions, duplication, and poor design choices, leading to rapid growth in lines of code and reduced human oversight. Traditional AI judges fail to assess “sloppiness,” while human review is accurate but unscalable; simple metrics like LOC change, verbosity, and erosion have shown promise for quantifying code quality.