Thank you for your feedback Vasiliy, we are glad you liked the article.

To your question: of course this solution is one out of many different possible ways to solve this problem. We were also thinking about other solutions, but we have decided for this solution, because it keeps the models small an clean and puts business logic about the JSON interpretation into one single place -> the definition of our parsing rules/Adapters.

The biggest advantage in this solution I think is that broken content can be filtered completely from our parsed data. If the JSON has no `cell_image` defined for the article, the whole article will become null (filtered). This may not be the useful for everyone, but it was our use case. However, we are still confident, that you can still build other use cases around this solution (see e.g. `VideoType.unknown`)

)