Academic literature is currently published and consumed in the PDF format so we are taking the approach that we should first augment the most widely used document medium.
The argument for posting academic documents as HTML on the Web is strong and should be done, but this relies on not only Internet connections being live but also on domains being active for generations to come. This is why we believe that the distributed method of connecting documents via citations should also be supported, not only web-links.
Similarly, there is absolutely nothing wrong with using web based or code based document formats (such as Jupyter Notebook) when the field in question allows it. However, if something is dependent on a server to be found it will become brittle over time. PDF documents can be widely distributed and read with any software so has a larger chance of long term survival we feel.
There is no reason why Visual-Meta cannot be implemented for the Web or in ePub in the future however, as an additional layer of metadata for accessibility and robustness.