◈mtknn @mtknn@cybre.space

experimenting with multiple colors + textured fills (the red fox skull 3d scan is from scan the world: https://www.myminifactory.com/object/3d-print-red-fox-skull-50865)

if you allocate a 3d texture on gpu but never upload data into it, and then treat this unallocated texture as a volume and raymarch into it, you get some interesting results

reaction diffusion, with f = 0.022, k = 0.059

so I keep a document of weird dreams I've had and I don't remember writing this *at all*:

6/20/2020:
Summoning dracula to fight him with the equivalent of yugioh cards
Dracula does not know anything about yugioh

here, have an art

still alive, just haven't... produced art I want to show in a while

move over 'synthwave' there's a new genre in town and it's called 'wizard disco' (ex: https://soundcloud.com/mildperil/paladin-arp-of-pan-final-version-2019)

I am going to capitalize and pronounce Latex (the typesetting software) like the word it obviously is and nobody can stop me

cooking tip: to convert ounces to grams in your head, simply multiply by 28.349523125

(and yes I'm aware there's an actual script variant Hershey font)

I love the Hershey fonts (http://paulbourke.net/dataformats/hershey/), so I thought: could I make a cursive version by just... not picking up the pen, and then blurring/smoothing the strokes

the result is a pseudo-font I'm calling 'machine cursive'.

I find these different kinds of "debug display" aesthetics fascinating

you can tell when a visualization has been created in opencv because it uses the Hershey simplex fonts (http://paulbourke.net/dataformats/hershey/)

for example, this video of tesla's autopilot: https://www.youtube.com/watch?v=5FITJg4KlMc&feature=emb_title

is a prime example of an opencv display, complete with Hershey fonts and rounded thick lines (because by default opencv draws thick lines with round endcaps and joins)

sometimes I look at the moon or the sun and try to convince myself to really believe (as opposed to merely believing in an intellectual sense) that it's a concrete physical object (as opposed to just an image or illusion or 'representation' hanging in the sky) and I don't think I've ever succeeded

"central planning is impossible because you'd have to solve a linear program with 12 million variables" sure is starting to feel a bit silly now that desktop GPUs are effectively supercomputers

(not that I'm in favor of central planning, but this line of argument is not nearly the slam dunk some people still think it is)

inside every random block of gpu memory lives a brutalist skyscraper

another view

a particularly attractive random slice of GPU memory

what's interesting is that the driver seems to want to give you the same memory block each time, so if you fill the memory with 0xFF, terminate the program and do other stuff for a while, if you run the program again and don't clear the memory you'll get a half-consumed block

if you allocate a 3d texture on gpu but never upload data into it, and then treat this unallocated texture as a volume and raymarch into it, you get some interesting results

it's hard to go wrong with bold color maps

mastodon is doing something funny to the gamma of these because they look quite different in the preview vs. if you click to look at full size

the algorithm "take a sphere and just subtract random other spheres out of it" produces surprisingly interesting structures