Version numbering policy (e.g. reason for missing 0.126.0 and 0.127.0)?

[your-diary]

I regularly check the release notes of codex, and often find some versions are skipped when a stable release is published.

For example, the next stable version after 0.125.0 is 0.128.0; 0.126.0 and 0.127.0 were skipped.

Regarding this, I have two questions:

• Just out of curiosity, why do you sometimes skip some version numbers? This is a bit inconvenient because I usually compare the latest version number with the local one to estimate how far behind I am.

• More importantly, when you release the version $N + M$ after the version $N$, skipping $N + 1,\ N + 2,\ \cdots,\ N + M - 1$, can we expect that the release note of the version $N+M$ contains all changes since the version $N$, not just the changes since $N + M - 1$?

[Boilerplate4u]

I understand exactly what you mean, especially the second point.

If a stable release jumps from version N to version N+M, it becomes hard for users to know whether the release notes for N+M include all user-facing changes since N, or only the latest internal/prerelease step.

I opened a related issue here: #21706. My concern is about Codex CLI release transparency more generally: clearer changelogs, useful prerelease notes, explicit breaking-change notices, migration guidance, before/after examples, and keeping documentation in sync with actual CLI behavior.

Please feel free to take a look and see whether it also covers your concern. If not, it would be useful to know what is still missing, because I think your point about skipped versions and complete release notes is an important part of the same broader release-process problem.