I tend to write my code comments in a very idiosyncratic way, and LLMs either don't understand that (nor do I expect them to), or it isn't much worth explaining. I'm fine with normal LLM comments, but publishing to production means I often have to do a lot of manual clean up with comments when committing code, and that's tedious.

It would even be fine if LLMs all commented the same, but they don't. I actually like the way some LLMs comment, but others are horrendous. The inconsistency is what makes this a mess, and simply saying "no comments whatsoever" (though it works) is hardly an alternative. What I want is a clean, consistent format with strict guidelines on what, and importantly what not, to comment, that works for all LLMs.

The idea here being that I can go on my merry way coding with agents, and that going forward they'll always output perfectly understandable code which I'll never have to check again.

Is there any progress here?