• How to Organize a Hackathon edit and release after THLI event on Jan 11
    • Part 1 will just be a list of resources to check out; very good resources have been written by much more competent hackathon organizers than myself.
    • Part 2 will be the meat of it. It will focus on how to figure out the purpose of your hackathon and designing a hackathon around that based on first principles, rather than doing everything the default way.
  • My answers to the most interesting questions from John Locke Essay Competition.
  • Lessons from failing at just about everything
    • USACO: not getting complacent + overconfident
    • Hackathon Organization: you need to put work in to create amazing environments, just getting cool people together doesn’t do that by itself most of the time + you don’t need to be the person running around putting out fires yourself, trust other people
    • Blogging: Stop being a perfectionist (not in the faux-humble job interview way) you’ll learn and write a lot faster if you stop editing over and over and over and linking everything to unnecessary degrees
    • AI: Don’t get scared of complicated-looking stuff
    • Club: Be less scared of feeling weird. When you’re recruiting and raising interest, if you sound like you’re embarrassed about your own club, that’s a signal that other people would be embarrassed if they were in it too.
  • A short post on productivity
  • The Tesla vs. Waymo comparison here nerd-sniped me. I want to read the relevant Tesla and Waymo research papers and learn a little more about sensor technology before writing about some of my thoughts.
    • Based purely on priors, though, I agree with Andrej Karpathy. It’s easy to say that Waymo is ahead based on what’s visible at a surface level, but Waymo has a hardware problem, which is probably going to be harder to solve than Tesla’s software problem unless the LiDAR issue just can’t be worked around (although apparently Tesla trains with LiDAR). Also, on priors, I agree with Ben Thompson that the history of AI so far has showed that you always want to bet on pure machine learning vs. encoding human knowledge if you’re looking at the long-term.
  • The explanatory power of not viewing large groups (like countries) in history as faceless entities with general motives and abstracted stories that attempt to explain their actions
    • This is nowhere near an original idea and something everybody knows, but I think it’s so easy for people to turn to general abstracted stories when thinking about history that it’s worth writing about
      • US works on hydrogen bomb because of competition between the different branches of the military (? I believe heard this on the Dwarkesh Podcast)
      • Soviet Union to Russia as a story of Yeltsin and Gorbachev hating each other
      • Weimar Republic to rise of Nazism as a highly contingent story filled with different factions trying to use the Nazis in their struggles against each other
      • Russia at the outbreak of WW1 as a story of Sazonov vs. Nicholas
  • On selection factors because I think it’s so important maybe looking at a lot of my beliefs, hobbies, values, identity, etc and analyzing them from this perspective
  • A follow up to the education reform post with a story about why things are broken in schools right now and how to address them.
    • Can probably digress into a case study of service-providing public institutions in general, why they’re so hard to execute well, and how to make them execute decently well.
  • Extensive Blog Post on Tools for Thought
    • Interesting work in general
    • Note-based work is overrated
    • Spaced Everything work seems to me to be incredibly underrated and under the radar
      • Spaced everything + note-based (just figuring out ways to make note-taking and spaced repetition work well seems underrated)
    • Search as tft feels underrated w/in tft places but definitely overrated (in the sense that I think too many people think they can “disrupt search” and are providing marginal improvements w/ no moats) in general
  • Breaking Talented High Schoolers out of Fakeness
    • The path of least resistance for most of these things is to repackage skills over and over again in different forms and do the bare minimum amount of “work” to get the maximum “results” a lot like the bs-ing at a lot of hackathons where you try to show the max impact and cool-sounding-ness with the least amount of actual technical work by sounding big brain enough
    • Things like Hack Club are an example of scalable systems that emphasize real learning what other things?
  • Chicago vs NYC (Manhattan)
  • Vibes-based:
  • A discovery fiction on something interesting that I want to grok deeply
  • Tensions draft near complete
  • What I want the future to look like
  • Search of the future
  • Building a Community for Tools for Thought
  • Discuss the technology in Making Massive Input Better