These factors lead the researchers to conclude that current AI coding tools may be particularly ill-suited to "settings with very high quality standards, or with many implicit requirements (e.g., relating to documentation, testing coverage, or linting/formatting)
What on earth? These are the projects that AI excels at because there are more, better examples to use from the code. When your code doesn't have high quality standards, you end up with multiple patterns to achieve the same thing instead. Similarly, high documentation or testing burdens means that it has more context and more guardrails built it, whereas a repo without these reqs wouldn't be able to verify changes as easily.