allison @aparrish@friend.camp

@tripofmice

@craigmaloney this is the worst thing I've ever seen

@tripofmice

@varx oh wait I just saw that you were replying to the original post, not my post downthread. the stacked distilbert states look initially look like the image on the left... once I subtract the mean, it looks like the image on the right (so, close enough to random noise, which is why I've been testing with random noise, but there are some interesting patterns and topography that will hopefully come through in the final plot)

@varx sure! step one is (as you guessed) grayscale uniform noise, step two is gaussian blur, step three is thresholding the gaussian blur at the image's median

@liaizon sure go ahead!

the colors are just for debugging (fading from red to blue according to the index of the given polyline)—the final piece will just be b/w. also this is using debug data (thresholded uniform noise w/gaussian filter), not DistilBERT states like I'll use for the final version. I'm at the point in the project where I need to break the code out into standalone modules instead of having one big spaghetti notebook

more progress! I ended up having to write my own skeleton tracing algorithm, which I thiiiink is working okay now—basically it does a flood fill starting with pixels having exactly one neighbor, and each connected pixel is a node in a graph; later I use visvalingam-wyatt line simplification on each segment between nodes w/3+ edges. this technique gives me nice long lines, clean intersections, & hopefully more elegant plotter gestures. (this is just a raster preview, will try an actual plot soon)

@nasser "nearest neighbors" from upthread feels like the right answer to me—you're resampling the array to be four times as long using nearest-neighbor interpolation. or maybe "zoom"? (since you're essentially "zooming in" on the array by a given factor?)

first attempt at sending this to the plotter. I ended up cutting the plot short because I could see some aesthetic and technical problems that I want to fix, and I didn't want to bother waiting another 45mins for the plot to finish, haha. but the basic idea is there and I think it looks nice?

this is a good article about the new structural pattern matching feature in 3.10: https://benhoyt.com/writings/python-pattern-matching/ demonstrating its benefits and discussing its drawbacks. the way that variable names and class function calls are "overloaded" inside of case statements is not super intuitive imo

@June this is the closest thing I know of https://github.com/dariusk/rss-to-activitypub/ but the category of people that would consider this "simple" is... pretty small

(darius runs a hosted version of this https://rss.friend.camp/ but it's only for friend camp users unfortunately!)

@cody yes... YES

the year is 2025. openai lobbies congress to force websites that publish user-generated content to guarantee content is free of synthetic data (and mark it as such w/metadata in the html). google lobbies for compulsory no-cost licensing of all content published to the web, unless the site owner follows [proprietary standard that costs millions to implement]. facebook pays below poverty wages to thousands of contractors in locked, device-free rooms to type sentences, any sentences as LM fodder