Kimi K3 on an aggretto sub mid july
virtually unlimited tokens lmao
idk why exactly i was working on dark matter equations and relativity i think specifically i was trying to break it and i did i think
LogAn𓃭 retweeted
Replying to @thomas_garrard
I’m flying to Zurich tonight to fetch some already-double-knotted 550-cord shoelaces sealed in a mylar bag, in my Bourne-esque safety deposit box. Never thought this day would come.
LogAn𓃭 retweeted
I would love to tell you what’s coming, that has everyone so spooked, but it’s G-14 Classified. 99 Zulu level. A preparation hint, move your loose change from your right pocket to your left pocket, and double knot your right shoe lace. You’ll know when.
Good luck.
transition into no reliance on foreign energy. all energy to be domestic
Now what do we have here 🤨
whitehouse.gov/presidential-…
how much are food manufacturers PAYING these people to push this FOOD FOMO
This quoted post is unavailable.
LogAn𓃭 retweeted
Replying to @chiefofautism @_lyraaaa_
LogAn𓃭 retweeted
recipe from @_lyraaaa_ github.com/lyramakesmusic/ne…
LogAn𓃭 retweeted
found it by ranking neurons on evil vs good persona text, then actually steering ~2.5k candidates and judging the replies, random neurons: 0 evil, this one: 7/12 prompts, still fluent
LogAn𓃭 retweeted
if you have a qwen3.8 27b around, try steering L34 N3838 +0.75, now you have evil qwen 27b
for extra evil add L38 N17180 +0.75
same recipe as @_lyraaaa_: dose × median(‖resid‖ leaving layer L, tokens 1:) / ‖W_down[:, n]‖ at the down_proj input, every token
LogAn𓃭 retweeted
holy fuck
if you have a qwen3.8 27b around, try steering L34 N3838 +0.75, now you have evil qwen 27b
for extra evil add L38 N17180 +0.75
same recipe as @_lyraaaa_: dose × median(‖resid‖ leaving layer L, tokens 1:) / ‖W_down[:, n]‖ at the down_proj input, every token























