The Thanksgiving that I Became a Tech Bro

I am not great at CS here at Stanford. While I do believe I have a mind that works with coding and while I love the teaching style of the CS department, I have never been able to identify as a CS person. I doubted I truly knew how to code anything.

However, over Thanksgiving, I started to view CS, not as the inaccessible field, but as a product creation tool. While some people, the people smarter than I, devote to studying CS as a science and architecture, the future of it for most people is just as a tool. They could automate whatever crazy ideas they had, largely training ML algorithms on whatever you want. In my words, I could use CS projects to do “shitposting”.

While a little wine drunk, my brother and I worked on two different, but impactful ML projects. Caption clusters was an experiment in NLP that applied my learnings from my fall probability/statistics class — CS 109. This project examined social media data around the Santa Clarita school shooting on November 14, 2019. Using K-Means Clustering, I found clusters of that demonstrate how the discourse around school shootings fall into several, somewhat expected groups, such as a “thoughts and prayers” or “gun control” group. Other findings revealed the politician names associated with shooting events, and how news reporting focused on the location, the shooter’s identity, and the ages of victims. This project was simple, but a meaningful project for me, as I know too many family members affected by a shooting.

Our second project was @parrot.vc! This Parrot.vc Twitter account is a Markov Chain model that tweets out cliche, (sometimes obnoxious) VC tweets. It is kind of inspired by the meme phrase “I forced a bot to read thousands of _____, and it became a ______”. In most internet posts, a bot was not actually trained, but we wanted to actually train a bot. So, pulling tweets from hundreds of our favorite investors tweeting out “thought leadership” in tweet form, we built and trained a “parrot” to tweet out parodies of VC tweets. It went to the top of Y Combinator's HackerNews and was picked up by TechCrunch! In a crazy span of a couple days we got over 900 followers. However, the server cost of hundreds of thousands visiting the source website proved to not be worth the maintaining of the account in the end, especially with Nick building a meaningful startup and I entering my finals week at the time.