• 0 Posts
  • 26 Comments
Joined 5 months ago
cake
Cake day: May 23rd, 2025

help-circle








  • this comment:

    Unless you’ve only ever used ChatGPT, you will know that LLM-produced code is not the result of a single prompt, not even a conversation, but rather a workflow that often goes as such:

    • Discuss a problem with the LLM. The LLM autonomously reads large parts of the repository you’re working in, during the discussion.
    • Ask it to write a plan. Edit the plan. Ask it about the edited plan. Edit it some more.
    • Repeatedly restart the LLM, asking it to code different parts of the plan. Debug the results. Write some code yourself. Create, rebase, or otherwise play around with the repository; keep multiple branches of potential code.
    • Go back and edit the original plan, now that you know what might work. Port some unit tests back in time, sometimes.
    • Repeat until done.

    is so stupid




  • The number one and number two most cited living scientists across all fields think scenarios like this are not only possible but likely to happen. And the average AI researcher thinks there is a 16% chance of AI causing human extinction.

    assigning a number to it makes it scientific

    aside/rant

    i wonder to what extent this bullshit works because of people’s fear of math

    i wish i could convince people that STEM skills are no different than a law degree, in essence — you’ll meet dipshits that are excellent mathematicians and you’ll meet smart people that are mediocre mathematicians. i suspect it’s because people view mathematical notation as impenetrable (when that just depends on the same shit any technical writing depends on, like the writer’s skill at communicating, the reader’s familiarity and strength with the prerequisite material, etc.)

    it’s frustrating, given the number of stupid mathematicians i’ve met








  • I guess there’s sway? none of these options entice me to be honest

    I used to use Sway. I found it tedious to configure several different things via config files. Kanshi in case you plug in a monitor, Waybar, Swaylock[1], etc. And, I may be misremembering, but you had to edit the Sway config to launch these programs at startup. There was just friction everywhere.

    I have been daily-driving COSMIC for about six months and it works pretty well, although there are infrequent crashes (less so since the beta release, I think). I like it as my tiling WM, but also occasional crashes don’t affect my workflow too badly.

    Wayland protocols are an almost ideal way to create intentional incompatibilities and network effects.

    Would you be willing to elaborate or follow up on this? I checked out the core protocol but think I’m way too out of my depth to relate it to what you wrote.


    1. Also there was a bug that allowed people to bypass your lockscreen by mashing keys. Sort of made me hesitant to try anything Sway again, although I believe the problem has been fixed. ↩︎