Gemini* and Gopher

• Personal/Opinions

  • Pandora's Box

    ~theoddballphilosopher

    Her first memory was lifting the box open, the sun blinding her eyes and gracing her tender skin.

    She had carried that box around her for as long as she could remember.

    It was all she had, the only thing that kept her bound to the past.

• Politics and World Events

  • Peace

    A Kentucky Party Executive Committee member who abruptly resigned this week first submitted some- thing beautiful I intend to publish at The Kentuckian[^1].

    My friends and I started the KYP to have a vehicle for political power to stop arming Israel, Ukraine, and wherever else this flailing empire might want to blow up.

• Technology and Free Software

  • re: emacs from scratch

    I had a similar story to Jack's for my vanilla emacs config. when I started out, I was simply curious to poke into the rabbit hole that is emacs. I had come from vim and neovim. I looked around, it seemed that customizing anything to be "satisfactory" (by my... overly high standards in retrospect), required a lot of elisp, and a lot of troubleshooting. to make matters worse, I had recently switched to MacOS, and was unfamiliar with where things were like linux, where things were inherited from BSDs and where things were completely foreign. Getting Evil to play nicely wasn't easy either.

  • I quit Drawabox immediately, 10/10 highly recommend

    I've written before about my fear of failure and my struggle to determine what it is I truly want to do. I had a funny experience with it recently.

    I've been considering taking up drawing as a hobby, so last week I went out and bought some fineliner pens and sat down to start Drawabox, a free online drawing fundamentals course.

  • How Determinantal Point Processes supports Out-Of-Distribution Generalization

    Deep Neural Networks are impressively good at doing human-ish things which until recently computers just couldn't do. However, one area where humans excel but DNNs are still terrible is out of distribution generalisation (OODG) - that is, applying knowledge learned in one context to another context which has never been encountered before. Now modern AI can sort of do this in practice, since presumably models like Claude and GPT-4 weren't trained on most of the prompts they get from users and they do okay. However, while we don't fully understand what's going on here, we do know that this only happens when DNNs are trained on massive amounts of data - a smaller model which is trained to be really good at one task isn't going to be good at a different task without training, even if the tasks are pretty analogous.

  • Internet/Gemini

    • Retrospective 0

      I neglected my little corner of the web for the last few days. I guess this is because the energy that I would dedicate to it goes into the advent of code challenge. This puts my arbitrary goal of 35 posts before January in jeopardy.

    • More “you should have your own website” propaganda

      I will say something potentially theoretically controversial, inasmuch as I don’t think I’ve seen anyone speak out against the practice: Dual-hosted Gemini/Web sites make both experiences worse. In Geminispace, bi-hosted posts are peppered with too many end-of-paragraphnotes or footnotes or what-have-you; on the Web they tend to lack even the tiniest bit of “you can do this thing that isn’t expressible in pure Markdown” flair.

  • Programming

    • Polyglot tomfoolery ⬜