Recurse Center: week 1 & 2

two weeks in new york city have flown by. here's a snapshot.

things i've done + learned

  • i built this dashboard for understanding the Twitter conversation around the upcoming Canadian federal election

    • i knew very little about web dev before this. over the past two weeks i learned lots about web development, data visualization, and politics Twitter

    • lesson 1: it turns out that it's extremely expensive to deploy a web app at scale which requires 7GB of in-memory processing, ha (i suppose i'm spoiled from working in machine learning, where high-memory, always-on servers are often the norm)

    • lesson 2: real-time Twitter dashboards are hard, because the number of favorites and retweets associated with a tweet are monotonically increasing. this means that tweet metadata needs to be updated daily, otherwise data is outdated (i.e.

      "top 10 favourites" may no longer be accurate). unfortunately there's no way to get around this without another pass through the data.

    • lesson 3: topic modeling on politics Twitter is extremely hard. i'm still working on this. the conversation is extremely complicated.

    • i sort of learned how to use Flask + gunicorn!

  • this is different than how i intended to spend the first two weeks, actually. i thought i would be focusing on deep reinforcement learning. however, misinformation + natural language processing is fascinating, and i think i'd like to focus on NLP going forward. i would still aim to fit in a study of deep reinforcement learning at some point, perhaps toward the end of my batch.


  • poking through the Twitter Elections Integrity Hub dataset on the misinformation operations related to the China/Hong Kong protests

    • i'm inspired by the Hamilton 2.0 dashboard, but it is unclear how difficult this will be to replicate at scale

  • understanding other ways that machine learning and natural language processing (especially large language models) can be used for misinformation research

other thoughts

  • there exists places in the world where everyone is genuinely kind and amazing and not trying to compete or undermine each other! i'm amazed by how RC has built a community of some of the kindest, brilliant, most interesting people. everyone here is working on ambitious projects and has excellent stories to tell

  • i now understand the totally-warranted cult following around Trader Joe's

  • i wrote about a coffee + bike tour we did this past weekend. it touches on cities, coffee, and all the magic of the third place

  • the best people in life are free

i love stories — write me a twitter dm 🔮✨

To reply you need to sign in.