Recurse Center: Self Directed Learning with a Community

— 3 minute read

It’s the beginning of Week 3 out of the 6 week half-batch at the Recurse Center.

If you’re unfamiliar, RC is a free, self-directed retreat for programmers to “work at the edge of your abilities, develop your volitional muscles, and learn generously” and that really describes what it’s like.

I’m remote but there’s a lot of opportunities to connect (and the option to opt-out). Daily check-ins are like stand ups. Self-organized recurring study groups are common. We even started a weekly in-person coworking session in LA.

I landed on a couple goals early on.

  1. Contribute regularly to PyTorch with the goal of contributing to a major feature.
  2. Learn the in’s and out’s of ML / AI and share learnings along the way.
  3. Join a team building cutting-edge ML / AI solutions to solve pressing and meaningful problems.

Contributing to PyTorch

So far, the strategy for contributing to PyTorch has been to start small. There’s a “good first issue” tag for people to start contributing. The goal is to get familiar with a small part of the codebase, build and run tests locally, and learn how PRs are submitted, reviewed and merged. Much like your first week at a new company.

So far, so good. I have a second PR approved and ready to merge - even if it’s a one line change. It was good practice on reading a Github issue, understanding what and where the issue was, and writing a clean change that would be accepted into the codebase.

Learning AI / ML

My first topic was the Transformer architecture. Given how much it revolutionized the deep learning space, it was a clear starting point.

I spent a lot more time than I estimated to clearly understand and then communicate the ideas behind the architecture. I thought it’d take 1 - 2 days. It took me closer to 5-7 days. It’s good to estimate beforehand, benchmark, and adjust your future estimates - much like day-to-day work at a job.

I shared it to LinkedIn and it got good reach and reception. It’s good to be in the zeitgeist if you don’t already have a lot of followers / distribution. Twitter was more like tweeting into the void.

Next

The new year has begun and I feel more motivated than ever. RC has given structure and peers that I missed having at a traditional working environment. I’m excited to continue learning and sharing with others to learn too. I believe we’re at the beginning of the next technological revolution - whether people compare it to harnessing electricity, breakthroughs from genome sequencing, or the creation of the database, having machines discover and decide for us will define the next decade.