plzdontkillus diary (#1)
first week of plzdontkillus is done. the first day i watched everyone's videos and concluded that i need to kill myself expeditiously. before this, i was basically at peace with the idea of short form video, and felt that the crusade against it is a moral panic. now that i have pondoored the topic more carefully, i say that short form is demoniacal, lacking in all virtue, and is essentially a planetary-scale egregore whose end is the reduction of humanity to humanesque cattle. short form by default is structurally hostile to conveying ideas with content (e.g. valid argument, evidence). it is for conveying vibes -- thin vibes, centered on the system 1, on appetitive instincts (e.g. for beauty, disgust, and tribal signalling). getting to see, in person, the vibrant and thoughtful fellows at pdku flatten themselves into their short-form-personas has been potent sui-fuel.
if you want a picture of the future, imagine a beautiful woman oiling up her feet to shoot a video about the anthropic ipo.
there are many good lads here. compelling talent one, two, three.
i don't blame anyone, but there is astonishing amounts of cynicism behind the production of short form video.
i see that there are three independent axes to work on:
- editing skills & taste
- virality
- ai safety / alignment messaging
it seems to me a very difficult problem to make short form videos that are all three (i.e. good, viral, and substantively about asi futures). tiktok is entertainment, and a particularly parasitical form of entertainment at that. alignment arguments are just a very poor fit for the medium. data center populist slop is way more virally fit.
so, optimistically, i can pursue at most two of these three goals. i am vacillating between 1/2 and 1/3.
i may also just reject everything that isn't goal 1. i have been just poasting my videos on X. i am currently more interested in building the reach of my twitter than finding an audience on tiktok/instagram/yt shorts.
one of my major online anxieties is that i am bad at tweeting. i believe i am just starting to build good tweet instinct. i think i will use my pdku videos as background material for growing my tweetoooor following.
i am making myself feel better by via effort justification. i'm trying to edit 8-13 hours a day. learning is fun.
my day 1 video was a jesus edit. i've watched an enormous amount of #jesustok, #catholictok. i knew going into the program that i wanted to try using diffusion models in new ways. i wanted to focus on the narrative in the gospel of mark (the most mysterious and austere of the gospels). however, since i wanted to take good paintings and prints and turn those into video clips, i was constrained by the parts of the gospel that had depictions i liked. so, mary singing the magnificat in luke is in, and so on. the editing is not hype enough (it's too much of a slideshow), but i find it satisfying to watch, despite its slopfeel.
my day 2 video was compute cop, which i think is actually good. the video is generated by creating a bunch of keyframe images (approximately 12), and then using seedance 2.0 to generate 5 second clips which interpolate between the two. i like the idea of making art videos that assume ai doomerist ideas as background value. i want to double down on this sort of hypermedia aesthetic.
on day 3 my capcut crashed and i lost like 6 hours of work (OOF). i crashed out and produced this cute video.
the actual video i was working on was a nick mullen shitpost. i posted a scrap of it here.
my day 4 video was usa 250, which is my take on the america-pilled hype edit. i am an america-maxi. unfortunately, i think american dynamism is going to create an asi-based apocalypse.
my day 5 video is a clone of a youtube short i watched yesterday that pissed me the fuck off. there's this slopmonger on yt shorts who takes frontpage reddit posts, voices them out in a disgusting voice, and edits cooking brainrot under it.
i will be doing a lot of video cloning this month to learn editing science. as expected, this was a very educational activity.