Whether or not that's reasonable in real-life [1]... learning the Twitter API + doing some sentiment analysis, etc is a good scope for what could be an interesting hackathon project.

You could run the trades on a simulated network or just print out the trades and calculate when/if the bot can beat the market.

1: Much automated trading these days is very high frequency, on the order of milliseconds (https://www.technologyreview.com/s/416805/trading-shares-in-...).