For people who query S3 buckets with very large numbers of objects, github.com/varveio/swath looks interesting. I haven't tested it (my needs are simple), and the author used AI/LLM code generation, but I spot-checked two details that give me confidence. The documentation recommends dynamically listing big buckets as a last resort and mentions AWS-idiomatic alternatives. More importantly, this is one of the few pieces of software I've seen that bothers to set the AWS SDK retry mode, a zero-effort way to increase reliability whenever we use an AWS SDK. (See the 2026-05-20 AWS Developer Tools Blog post, "Announcing updated retry behavior for AWS SDKs and Tools".) I like the way the author used AI not just to generate code but also to explore how other parallel S3 bucket listing tools work. A readily apparent drawback of code generation is vastly increased complexity; there are too many files and too many lines to read. Everything rests on whether comprehensive distributed systems testing is possible and whether the tests have caught potential race conditions.