AI-generated changes are easy to create and hard to review.

AI-generated changes are easy to create but difficult to review because large tasks often mix unrelated changes into a single working tree. A better workflow would preserve logical slices independently. Meanwhile, unsupervised AI agents writing Flutter code frequently commit structural failures, such as poor state management and missing tests. One example involves an agent using a subprocess call to curl instead of standard libraries like package:http or dio, making the codebase expensive to maintain.