Code changes have at least a linear amount of bugs on average, but have a large random component and selection bias - A key part of being a good developer is le

Code changes have at least a linear amount of bugs on average, but have a large random component and selection bias - A key part of being a good developer is learning how to read code and notice/realize when it does have a bug. Even if the number of bugs is largely random each change set, repeatedly selecting the most obviously buggy code to change will reduce the number of bugs over time. This can apply at a number of different points and scales: The developer writing the code for the first time and testing it locally themselves, the code reviewer looking at the code, other developers reading and editing the code while working on their own features, other developers choosing which library to use for a given function and which projects to support, customers choosing which product to use and business to support. Each of those layers selects against bugs over time, eventually bringing it sub-linear to the end users.