🏛 Stoas near [[@agora/2022 11 26]]
📖 HedgeDoc at https://doc.anagora.org/2022-11-26
📖 Etherpad at https://stoa.anagora.org/p/2022-11-26
📹 Jitsi at https://meet.jit.si/2022-11-26
📚 Node [[2022-11-26]]
↳ 📓 Resource [[@agora/2022 11 26]]

Saturday, 11/26/2022


I've been thinking about winning. I love feeling the drive to win. I want to be able to summon up some feeling, some content that puts me in a particular state of mind… that gives me the determination to do something impactful and important. A motivational speech or something… how valuable would this be? How do we productize motivation? The ability to motivate everyone - as I suppose some content creators attempt to leverage traditional socials for - would be beautiful. (The other implications of such a technology - invoking compassion, revisiting loss, experiencing love - are just as valuable).

I've been seeing a lot of this content lately. Sports, and any other athletic work (as opposed to knowledge work) needs motivation; you need to be the best in the world at that moment, or at least good enough to beat that other team. Jamaal Williams speech from Hard knocks touched me recently… I didn't care at all about football until I saw him feel so strongly about the Lions - and so passionate about his position and his team. Likewise, the news that George Hotz has flipped from Comma AI (a company that aims to both "make driving chill" and "win self-driving cars") to Twitter after Elon's takeover with a desire to revolutionize the platform and impact what's most important, has had me flip through some writing of his discussing his - and, by extension, the Comma - philosophy.

Key takeaways:

  • Real life is the best game anyone can play. To me, this contrast is like living in Coq and writing POPL papers vs. freeballing with Common Lisp or JS at a startup. Video games, sports and type theory have simple, beautiful axioms, whether they're "get the ball into the goal with no hands" or the "Calculus of Inductive Constructions", and allow players to derive and develop complex strategies from them, often - in sports in particular - allowing the exploration of those constraints to evolve the game itself (with patches of an exceptional nature, similar to the law; Preston [enlightened me](https://twitter.com/prestonattebery/status/1596653459721576449 ) about the depth of the game of basketball here, for example).

    Those ideas are beautiful, and they derive some ideas from reality to approximate it - but, notably, they aren't. My stack of Coq proofs failed to consider that someone pouring water on my verified PCB would take my lightbulb machine out. Someone recently pointed out to me that we tend to over-correct or over-assume that skills we develop under different models will extend or apply to others, often in an attempt to justify that perfecting an easy task will help perfect a difficult one, but this doesn't work. Tasks done in direct support of the real deliverable - say, practicing shooting on goal to hone your shot accuracy for a real game of football - will pay off, but unless performed deliberately in support of an end, some means like squatting at the gym to focus on hamstrings may not actually correct your poor footwork in the penalty box if that problem isn't your biggest blocker.

    A lot of software systems will let you fall down this trap, too; fighting the typechecker to justify things is probably the biggest example of this today. In "Stop Writing Dead Programs", Jack Rusher hammers this home better than anyone else. Types can construct games that can be played, but they don't construe real, runnable code; you're justifying your work against a pretty abstraction of reality that lives in type theory, not the practical world that your product occupies. His thesis, which I echo here, is that improving the "liveness" of your system is the biggest improvement you can make to your development process. Getting end-to-end testable and runnable code into the real world as soon as possible, testing by running against a real system with real people and getting feedback, is the only thing that matters; otherwise, you're playing the typechecker's game, the abstractor's, or the reasearcher's rather than testing against reality. You want your code to effect real change in the world - so you should test it that way too.

  • Make knowledge freely available to encourage progress and competition. Comma AI is completely open-source, robust software, built for anyone to use to augment an existing tool; this is central to its value proposition. Any auto manufacturer or wannabe self-driving car startup can download Comma, throw it in their vehicle, and ship a "Powered by Comma" (MIT license abiding) self-driving system of their own. Hotz recognizes the significant surplus of value that's given away and doesn't really care - everyone should have access to the knowledge, and he has the confidence that his team will directly outcompete anyone forking and selling his stuff. if someone kills them at self-driving, they deserve to win, and the mission of true self-driving cars is accomplished anyways; company finances are secondary.

    Likewise, if my car isn't supported by Comma, I'm free to add support to its control system to the software to make it work. The job of the company is to develop the technology necessary to allow cars to drive themselves; if I want it, my job is to make my car drive itself. I can buy the devkit and make it work. Dozens of people have. The ability for anyone to fix anything creates demonstrable value for the company and for people. Adding my car isn't some one-off hack; it's a way to let anyone with my car use this self-driving system without additional work. Everyone wins.

  • Solve only the problem you need to. Comma deliberately applied specific constraints at the outset: their product is, and has always aimed to be, a single plug-and-play hardware box that augments your car without modification. The entire system has been developed with this constraint in mind.

    Other companies all hijack their hardware in some way; some make design affordances in existing vehicles and play the game of car collaborations to weasel their infrastructure in, others happen to be the car company, and one company is building their own car from scratch to make the thing happen.

    Guess whose software benchmarks best?

The importance of people - through their writing, words and actions - to motivate others cannot be underscored. Blue Lock - at its face a manga about soccer - has been particularly motivational for me; whether this is wholly the intention of the author, I'm not sure, but my tenure as a mediocre soccer player from 6-18 absolutely plays a role (my relationship with the sport was something I'd always done).

Sure, games aren't real life, but the way in which Kaneshiro distills the sport into its motivating factors with the deliberate intent of imparting inspiration is fascinating. I pick up a chapter, watch Yoichi kill it in another match by deliberately observing his faults, then looking at what others do to improve himself and win, and feel like I can be the best at something. I'm not going to win self-driving cars like Hotz, but I'll be fighting to improve the world of agriculture - and, in the meantime, I'll become a damn good produce developer.


Receiving pushes... (requires JavaScript)
Loading context... (requires JavaScript)