Current classifiers are naive token-based implementations of basic ML models. Need to improve accuracy by incorporating more NLP steps and better validation

add price module using an api, this one looks good https://www.cryptocompare.com/api/#introduction

Rendering all the data makes the database sad

• Add elixir pagination dep https://github.com/drewolson/scrivener
• Show 20 tweets per page

use box or card to mimic the look of twitter tweets

• hit /bitfeels/api/stats for stats data (on page load)
• populate plot with data

get stats from db to d3

• pull stats from db
• make stats endpoint
• render stats as json

show database stats under bitfeels header

• number of tweets
• number of feels
• number of users
• number of stats

we have more data than we're currently displaying

show things like twitter handle, favorites, retweets ...