I don't get the thing with automated PR and commit messages. They mostly describe _what_ was done and not why.
Also, if you enhance your available information with AI, why not just write no message at all (or a short one as always) and use the AI when _looking_ at the commits/PRs?
The AI will certainly be better at this in a later stage because they themselves will get better and they will have more context due to being able to look at the commits that followed the PR/commit.
There is no point in generating the commit message when committing.
When auto-generated messages will become a commodity, there will be issues with it. For example, if the message doesn't fit with the actual commit contents (semantically or syntactically), I have to think hard whether this is an AI-message-generator-bug, or the author missed something or wether I am missing something. This is not cool and makes reading commit messages harder.
Also, if you enhance your available information with AI, why not just write no message at all (or a short one as always) and use the AI when _looking_ at the commits/PRs? The AI will certainly be better at this in a later stage because they themselves will get better and they will have more context due to being able to look at the commits that followed the PR/commit.
There is no point in generating the commit message when committing.
When auto-generated messages will become a commodity, there will be issues with it. For example, if the message doesn't fit with the actual commit contents (semantically or syntactically), I have to think hard whether this is an AI-message-generator-bug, or the author missed something or wether I am missing something. This is not cool and makes reading commit messages harder.