Thomas

17 posts

Thomas

Thomas

@DomaOrut

Katılım Mart 2026
35 Takip Edilen0 Takipçiler
Thomas
Thomas@DomaOrut·
@BenShindel @nerdosyndical where he gets money doesn't change that he is making a vibes based vacuous value judgement, which is doubly ridiculous because he comes from the most coddled generation in history of this planet. you guys need to grow a thicker skin when it comes to hearing hard truths.
English
0
0
0
20
Ben
Ben@BenShindel·
@nerdosyndical NYU Stern, which pays Haidt’s salary, is one of the best business schools in the world and famously is particularly good at career prospects for its students. This is a completely silly line of attack. Getting an MBA from Stern is almost certainly a positive financial move.
Ben tweet media
English
1
0
0
55
Thomas
Thomas@DomaOrut·
@natolambert @reflection_ai poor choice of words on my part, nothing wrong with optimistic and cheerful takes. you never shit on anyone and this was pretty brutal.
English
0
0
3
579
Nathan Lambert
Nathan Lambert@natolambert·
@DomaOrut @reflection_ai Yes, a big knock “mild mannered and positive”. I just want open models to actually work and exist and not be a branding credit
English
1
0
42
3.1K
Thomas
Thomas@DomaOrut·
@LittleKnownPL @YIMBYPoland Smartest person in Poland recently looked up the word nuance and learned this one trick to shove it between "black and white"^
English
0
0
2
57
LittleKnownPoland
LittleKnownPoland@LittleKnownPL·
@YIMBYPoland >O-1 visa itself indeed is not sufficient enough to be exempt? I never said anything to the contrary. And that's not what your initial post claimed. I mentioned "nuance", while you talk black and white.
English
2
0
1
1.1K
legal YIMBY alien
legal YIMBY alien@YIMBYPoland·
Translation: if you’re a advanced researcher at OpenAI on the O1 visa (extraordinary ability) and you apply for a green card, they’ll make you go back abroad and wait years for an appointment. Of course, by then you probably no longer work at OpenAI.
Daily Caller@DailyCaller

EXCLUSIVE: Trump Admin Closes Loophole Letting Migrants Stay In US While Awaiting Green Cards: 'We're returning to the original intent of the law' dlvr.it/TSgK6R

English
92
76
2.1K
261.4K
Thomas
Thomas@DomaOrut·
@stanislavfort you can just not include a paper you didn't read into citations then. the authors of related papers will reach out themselves to make sure you do lol.
English
0
0
0
27
Stanislav Fort
Stanislav Fort@stanislavfort·
Lenka is correct here, Hunter is wrong. This is not a big deal. There are different kinds of citations in a paper on a spectrum from "I actually used this result in detail" (you read those carefully), all the way to "some people did something similar in this space" (you don't really read those in detail because they don't matter much for what you're writing anyway). What is the issue here exactly?
Hunter Ash@ArtemisConsort

Academics are now openly defending the practice of citing papers they haven’t read.

English
167
7
237
110.8K
Thomas
Thomas@DomaOrut·
@natolambert @reflection_ai not a knock against you, but you're usually so mild-mannered and positive that seeing a dunk like this is saying something about these guys lol
English
2
0
13
3.8K
Nathan Lambert
Nathan Lambert@natolambert·
@reflection_ai i want y'all to succeed but it's a little self-serving to say "America's open source AI lab" having released nothing
English
16
4
792
48.6K
NewsWire
NewsWire@NewsWire_US·
Virginia Gov. Abigail Spanberger Vetoes Bill to Restrict ICE Agents at Courthouses, Schools and Hospitals
English
439
459
3.9K
2.6M
WolfSev
WolfSev@wolfsev1·
@HeverCastroB Lo dices como si la CIA no torturara a inocentes
Español
1
1
9
4.7K
HevercastroB
HevercastroB@HeverCastroB·
El presidente de China Xi Jinping tiene una sola hija, y no la envió a la Universidad de Pekín o alguna institución de prestigio en China, ¿saben a dónde la envió en el 2010?. 🤔 A Harvard, en el imperio mismo, con un nombre falso, así son todos los comunistas, aman al imperio.
HevercastroB tweet media
Español
836
7.1K
22.2K
902.7K
Thomas
Thomas@DomaOrut·
@zephyr_z9 you're usually pretty sharp but this is dense because you're assuming they couldn't have done that to India's patchwork cybersystems with existing resources already. they sell all of their military equipment preinstalled with exploits. doesn't mean Mythos won't wreck China's shit
English
0
0
0
60
Thomas
Thomas@DomaOrut·
@Davis58A dude has one dialogue tree pathway that he applies to random women because he is an NPC retard
English
1
0
6
165
Thomas
Thomas@DomaOrut·
there should be some kind of LLM speedrunning competition where you straight up encourage the model to solve things through the quickest shortcut
English
0
0
0
78
Cüneyt
Cüneyt@Cuneyt_S_Sevi·
Liberals on twitter will really be like "you believe in systematic change? that pales in effectiveness to my strategy, winning an election" and then not win an election
English
184
1.9K
16K
240.1K
Thomas
Thomas@DomaOrut·
@AryamanIyer3 @cHHillee man that hasn't verified anything in his life, including the output of the LLM bot responding on his behalf^
English
0
0
0
16
Horace He
Horace He@cHHillee·
A conversation I've had several times lately: Them: "LLMs are so good for learning whatever you want — there's never been a better time to learn!" Me: "When was the last time you learned a new domain with LLMs?" Them: "Oh, there's no point. LLMs do a better job than me already"
English
33
11
471
37.2K
Saurabh Kumar
Saurabh Kumar@drummatick·
The bar is on the ground cause if you’ve done even ML101 from any free resource you can answer all questions For 1, you just take first derivative, equate it to zero and get mean as the optimal assignment for centroid Same for 2, just take double derivative and show the sign is not always positive or negative.
Reso@Resorcinolworks

Interview experience at Applied Sciences-2 @ Msft questions are focusing more on mathematical proofs and derivation now.

English
14
22
440
50.8K
Alexander Doria
Alexander Doria@Dorialexander·
@boochi_dot_dev Train the base model with structured data already (which Moonshot has done, as any proper synth pretrain)
English
1
0
7
1K
Alexander Doria
Alexander Doria@Dorialexander·
Perplexity-based evals? In 2026?
Aman Sanger@amanrsanger

We've evaluated a lot of base models on perplexity-based evals and Kimi k2.5 proved to be the strongest! After that, we do continued pre-training and high-compute RL (a 4x scale-up). The combination of the strong base, CPT and RL, and Fireworks' inference and RL samplers make Composer-2 frontier level. It was a miss to not mention the Kimi base in our blog from the start. We'll fix that for the next model.

English
6
5
139
26.3K

Keşfet

@BenShindel @nerdosyndical @natolambert @reflection_ai @LittleKnownPL @YIMBYPoland @stanislavfort @StayingPoor