Curating massive music libraries. I’ve been using a small embedding model to organise my music for DJing, and being able to generate a t-sne plot clustered on perceptual similarity has been wonderfully useful.
I’ve also found CLIP models useful for searching videos, just embed a screenshot every couple of min of footage and query with a description of the scene.
And as bad as generated subtitles can be, when the only other option is nothing at all they are pretty nice to have.
A little bit of water too, as a treat.
I’ll always be thankful that vaping got me off tobacco, and eventually nicotine too. But there’s an abyssal gulf between the self assembled, diy vape and self mixed liquid that vaping started at, and the white ox grade addiction pods that vaping has become.
There’s absolutely no reason to have the default strength be 50mg/mL, unless optimising for addiction without quite hitting a poisonous dose.
Chugging a quad shot espresso while saying ‘get up trinity. Get. Up.’ in my head
If you time it right, the anxiety from the caffeine hitting really sells the feeling that a matrix agent is coming to kill you so better get moving.
Torrents-csv.com is pretty nice too.
Topologically speaking, no. This is clearly a sphere.
Makehuman for generating models, unity preview renderer running as an MCP tool, language model reads page instructions for pose info, low frequency noise on pose bones to add micro movements.
Bots have had a workaround for weeks now. Same with browser movement bot detection, just train a behavioural cloning model on human mouse movements, and use that to create mouse trajectories instead of directly moving the cursor to the target position.
Cringe is a spook of social conformity, kill the cringe cop in your head.
Kill it Le epic style like the wholesome chungus 💯 you are
Gentoo, patch it out, funroll-all-loops for the memes.
I’ts open source, change the shit you don’t like.
Opiates had evidence of reducing the risk, so I guess the lesson is have speedballs if you want speed
Abandon git, return to subversion.
Bloodmouths and cheese breathers should be factory farmed
No idea, I just have the maths autism
Abandon axis ideology, embrace k-means tsne political clustering
With a large enough tangent kernel and histogram equalization, every monotonic nonlinear relationship can be projected into a series of linear decision boundaries
They’re proud capitalists, so it never bothered them. They aren’t proud of their slopware though
If you want more nuance, you need more dimensions to display them. The compass is basically taking two principal components of something with arbitrarily high dimensionality, in order to flatten this multidimensional conceptual ‘political ideology’ position to a plane .
The easiest way is to add planes with axes that are orthogonal to the left/right auth/anarcho axes, or alternatively you can decompose left/right and auth/anarcho into their own 2d components.
It’s not the better model; it’s the framework to build a better model. My initial ideas are using a second plane with enviro/industrial and tech/primitivist as a secondary plane. They aren’t completely orthogonal to the main axes, but should be different enough to get new decision boundaries.
CAD files and a BOM.
Ouchie, got me right in the TCP stack
I’m going all-in on transonic fluid dynamics. All the joy of nonlinear acoustics, with a little navier stokes as a treat.