anyone here fine-tuned a local model on game rewards? want to teach my qwen to chop trees faster than 125 xp/min
thinking lora on mac first then rl somewhere with real gpus
my animation linter says the owlet takes 14 steps a second when he walks. and when he runs
i watched it like 50 times and never noticed
i imagine any neurons activations if viewed in the correct perspective will render its purpose immediately and irrefutably obvious
wrote a linter for animation. it watches every bone at 60hz from the game camera and flags pops, foot slide and jitter
first run: 193 flags. the frog's feet slide the entire time he runs
mosswick retweeted
I know we’re past the “read the code/don’t read the code” argument.
But if someone can come up with a human consumable representation of the code which is right in the middle of lossless raw code reading and lossy description by the agents, I am willing to spend a fortune.
this interp thing is confusing because the first example of polysemanticity being incorrect is like the most obvious fucking thing im actually mad about it
when you preemptively solve problems people will sometimes seem offended, in reality its moreso they do not know why you do the things that you do, and that confusion and shame comes off as anger, frustration, etc.
shame is probably the most dangerous human emotion
its so easy working in not sf/nyc because what the fuck is a technical lmao who is leetcode never heard of em i click that meeting link ask how their weekends looking and make a few jokes and they offer me the job every. single. time.
