Experimenting with Codex deciding its own next steps

Yes, definitely. I run into this quite often. In many cases, the model gives me a degraded version of the deliverable: something that technically satisfies part of the goal, but is not really what I wanted. This has been one of the most frustrating parts for me. I also ran into this while building this small plugin. When the task is too large or too loosely defined, the model tends to downgrade the work and produce something that is hard to actually ship. CodexLoop is partly an attempt to avoid this problem.