On the API we don't make a strong (typed) difference between code cell and markdown cell - we only have the CellKind-hint. That makes some things a little cumber some, e.g

• Implementors of executeAllCell need to make sure they don't execute a markdown cell. Should we have notebook#codeCells or notebook#getCell(filter: CellKind)?
• One can assign output to a markdown cell which I believe isn't wanted
• We always allow to change the language of a code cell to markdown, which at least for GH notebooks is unwanted

So, I wonder if notebooks should draw a more crisp line between code cells and markdown cells so that they are harder to confuse and more restricted in what you can do with them.