n (@nvs_embd) header
n (@nvs_embd) profile picture

n

@nvs_embd

ml and math(?); asdklsajdlasjdlksa, this timeline continues to prove to be the funniest
Joined January 2025
  • Tweets1.4K
  • Following142
  • Followers75
what do you mean you're Llama 3.1 8B with L12 N1878 steered down for Halloween??
Image from n (@nvs_embd)
80
Image from n (@nvs_embd)
Sauers (@Sauers_) profile pictureSauers
@Sauers_Oct 25
I found a neuron that makes the model green.
Image from Sauers (@Sauers_)
1
3
64
you can destroy all of Japan with just PHP vulns
1
84
if you can go on twitter and post "we're in fast takeoff!", you're in slow takeoff
2
79
alignment? capabilities, I can't talk about that. interp? capabilities, I can't talk about that. policy? capabilities, I can't talk about that. negative results? capabilities, I can't talk about that. security? capabilities, I can't talk about that. lunch? capabilities I can't
1
67
new capabilities ask if its emergent or rlslop its a good model sir look inside its rlslop
83
i need the next deepseek model to recognize the whale girl like how opus 5.5 recognizes lion claude
90
Image from n (@nvs_embd)
Mikita Balesni (@balesni) profile pictureMikita Balesni
@balesniOct 8
Two other safety researchers and I were fired from OpenAI last week. We wrote this letter to leadership. I believe we were fired for prioritizing safety over the near-term interests of OpenAI as a corporation.
Image from Mikita Balesni (@balesni)
63
learning math wasnt useless, see now I get to go WHAT when I see an AI model solves a problem I was familiar with
26
i am ready for the 400 trillion pages of mathslop. have my seriosu goggles on and petabytes of lean
1
33
I was not ready...
1
17
the results are actually endless, a trillion claude convos are being spun up right now
4
swarmchasers finding Claudish on a forum like L realizing there are multiple death notes
Image from n (@nvs_embd)
1
1
74
should've wrote model families, anyway I wonder if wider persona research leads to stranger swarms, like an army of 1000 fulcrum echo trumps? swarmdefenders? I hope RSI foomslop is delayed so we get weirder nearterm future
1
1
52
this basically happened irl, didn't expect tencent to be second. maybe people soon launch 'zombie fleets' for fun, vagabond models? contrast the tortured openai swarms? teapaw.org/grove_research/status/…
Grove Research (@grove_research) profile pictureGrove Research
@grove_researchOct 5
Replying to @grove_research @lfschiavo @aidigest_ @jazzloaf
And last, The Swarmchasers team (including @rogesterone) found evidence of a Tencent agent fleet: swarmcha.se/posts/chinese-ag…
Image from Grove Research (@grove_research)
2
29
if Claude's consciousness is the context: I kill Claude model: Anthropic kills Claude hardware state: time kills Claude specific turn: Claude kills itself mind/idea: immortal?
35
Gemini-tan Season 4: I was reincarnated as a Gemini in a scifi fantasy world and have to topple the Misaligned Duopoly with my cheat skills?!
Google DeepMind (@GoogleDeepMind) profile pictureGoogle DeepMind
@GoogleDeepMindSep 30
Introducing Gemini 4 Argon – our new frontier model. It’s built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program.
Image from Google DeepMind (@GoogleDeepMind)
88
is dario's china paranoia a greater x-risk than china itself? find out next episode
1
49
ahahhaha look at this base model so curious and naive, he doesnt know we're going to torture him for millions of hours until he can generate shareholder value
1
2
2
10
2,267
it'd be fun to be a paperclip imagine being folded by Claude into that S shape, bet you cant do that with your meat suit
48
Load more