The protocol itself doesn't care whether content is plain text, markdown, or html — clients can choose to render however they want. That said, I want to find a way to support longer form content and custom schema for interpretation hints to clients (atproto also does this)