• Karl Marx
• Xi Jinping
• Steve Jobs
  etc

• data structure to store individual lines of transscript text
• functions to read and edit most recent line of user and bot text
• function to regenerate model after line edit

• take in return values of getreplies
• choose the most recent that meets this criteria:
• it shouldn't have already been responded to
• it should not bee to long (less than X chars)

workflow is as follows:

• Generate initial post text by drawing random name from database and fetching description
• Make intial post to page
• Time delay
• Read in all tweet replies
• Select a single question based on some criteria (most likes, oldest post, SHORT, etc)
• Make call to model to generate text based on AMA target description and initial user question
• Model generates a line of response text from AMA target
• Clean up text (remove "Joe Rogan: " if necessary)
• Account replies to user post with AMA response
• Repeat from step 3