There's a few really awkward corner cases that are much clearer in YAML (and, it seems, LCON). The most significant one, to me, is lists of objects, which are perhaps not so common in code literals (hence why people don't tend to run into it in coffeescript), but do tend to come up a lot in structured documents. In JSON they are trivial:

    [
      {
        "a": "b",
        "c": "d"
      },
      {
        "e": "f",
        "g": "h"
      },
    ]

In YAML also trivial and very concise:

    - a: b
      c: d
    - e: f
      g: h

But in CSON (if I remember correctly), I think this looks odd and confusing by comparison:

    [
      a: "b"
      c: "d"
    , # <- note dedent, without this it comes out as one object not two
      e: "f"
      g: "h"
    ]

Now, in each of these you can just revert back to JSON for the difficult part, but that kind of defeats the purpose. I think YAML-the-syntax strikes a really nice balance between eliminating unnecessary punctuation and having solid structure, and LCON seems like it does as well. CSON errs too far on removing punctuation without a lot of thought put into the remaining structure.

If I understand correctly, the above example in LCON would be (at most minimal):

    . a b
      c d
    . e f
      g h

I actually really dislike this use of . and would like to know a rationale for it, to be honest.

This project was partially a response to CSON. CSON was almost (but not quite) good enough for my purposes, so I made something that was.