The Lines of Code Fallacy: Quality Over Quantity in the AI Age
The tech world is buzzing with another bold claim about AI coding assistants. This time, it’s about Cursor apparently churning out a billion lines of “accepted” code daily. Reading this while working on a legacy codebase migration project at my desk in South Melbourne, I had to pause and reflect on what this really means.
Numbers can be deceiving, especially in software development. Back in my early career days, I remember the misguided pride I took in writing hundreds of lines of code daily. Now, with decades of experience under my belt, I take far more satisfaction in deleting unnecessary code than adding new lines.
The discussion around this claim reveals a fundamental misunderstanding about software development. Having spent countless hours refactoring codebases, I’ve learned that measuring productivity by lines of code is like measuring a chef’s skill by the weight of ingredients used. It completely misses the point.
Recent studies have shown that developers, regardless of their employer or salary level, typically write around 100 meaningful lines of code per day. This isn’t a limitation - it’s a reflection of the thought and care that goes into crafting maintainable software. Just yesterday, I spent three hours reducing a 200-line function to 20 lines, making it more efficient and easier to understand.
What’s particularly concerning about these AI coding assistant metrics is how they might influence management perspectives. Look at what happened at Twitter when a certain billionaire tried to evaluate engineers based on lines of code - it was a disaster that any seasoned developer could have predicted.
The real value of AI coding tools lies not in their ability to generate vast amounts of code, but in their potential to handle repetitive tasks and provide helpful suggestions. They’re most effective when used as assistants rather than replacements, helping developers focus on the architectural and design decisions that truly matter.
The software industry needs to move past these simplistic metrics. Good code isn’t about quantity - it’s about clarity, maintainability, and solving problems efficiently. Jeff Atwood said it best: “The best code is no code at all.” Every line we write is a line that needs to be maintained, debugged, and understood by future developers.
Looking ahead, I hope we’ll start focusing more on meaningful metrics: How much technical debt did we eliminate? How many bugs did we prevent? How maintainable is our codebase? These are the questions that matter, not how many lines of code an AI can generate in a day.
Let’s celebrate the developers who write less code to solve problems, not the tools that generate more code than we know what to do with. After all, in software development, less is often more.