skybrian @skybrian

Accordion player, former software engineer. In favor of rounding towards uncertainty, complicating the narrative, and thinking ahead. @[email protected] skybrian.substack.com California Joined August 2008

Tweets

3K
Followers

123
Following

180
Likes

121

Tedros Adhanom Ghebreyesus @DrTedros

4 weeks ago

Today @WHO supported the medical evacuation of 32 child and 6 adult patients from #Gaza, alongside 99 companions: * 25 child and 6 adult patients were evacuated to Italy * 5 child patients were evacuated to Belgium * 2 child patients were being evacuated to Türkiye We thank…

27 97 196 43K 4

Download Image

Michael Kofman @KofmanMichael

a month ago

An update on the war following a recent trip. As Russian offensive presume mounts, the front is not at risk of collapse, though salients have formed. More concerning is that Russian improvements in drone employment have reduced Ukraine’s advantages. Long thread. 1/

60 880 4K 660K 1K

Download Image

Omar Khattab @lateinteraction

4 months ago

DSPy's biggest strength is also the reason it can admittedly be hard to wrap your head around it. It's basically say: LLMs & their methods will continue to improve but not equally in every axis, so: - What's the smallest set of fundamental abstractions that allow you to build…

DSPy @DSPyOSS

4 months ago

32 58 840 392K 535

48 149 982 333K 1K

Nichole “Vapor” Ayers @Astro_Ayers

2 months ago

I’ve received a lot of questions and comments about this picture. It’s a testament to the unique view we have from the @Space_Station and what we can learn about the Earth from space. Also, turns out, this is a gigantic jet, another type of TLE. So cool to learn as we go up here.…

39 95 511 27K 71

Download Image

Zack Witten @zswitten

a month ago

Zack Witten @zswitten

2 months ago

https://t.co/VqkRK8aaYJ

9 29 1K 323K 333

Download Image

24 283 5K 723K 963

Download Image

Neel Nanda @NeelNanda5

3 months ago

Excited to have supervised these papers! EM was wild, with unclear implications for safety We answer how: there's a general evil vector. Boosting this is A solution to SFT on any narrow evil task We don't know WHY it's so general, but release better EM models to boost research

Ed Turner @EdTurner42

3 months ago

16 50 263 73K 207

Download Image

9 15 190 18K 95

Keenan Crane @keenanisalive

4 months ago

Fun new paper at #SIGGRAPH2025: What if instead of two 6-sided dice, you could roll a single "funky-shaped" die that gives the same statistics (e.g, 7 is twice as likely as 4 or 10). Or make fair dice in any shape—e.g., dragons rather than cubes? That's exactly what we do! 1/n

12 92 783 58K 325

Download Image

Patrick McKenzie @patio11

4 months ago

This week on Complex Systems, an in-depth exploration of far UVC with @JacobSwett of Blueprint Biosecurity. Far UVC is an ultraviolet lighting technology which may be as important for infectious disease control as water treatment and food supply chain management before it.

10 16 106 18K 16

Science Banana @literalbanana

5 months ago

I don't care if it's true I just love it

10 81 1K 58K 158

Download Image

Curiosity @MAstronomers

5 months ago

The line between day and night as seen from Space

59 678 5K 307K 331

Download Image

Michael Kofman @KofmanMichael

6 months ago

An update on the war following a recent trip. The situation has improved compared with Fall 2024. Russian offensive momentum slowed significantly over the winter, though it is premature to claim that the front has stabilized, especially following AFU withdrawal in Kursk. /1

68 1K 7K 1.2M 1K

Apollo Research @apolloaievals

6 months ago

AI models – especially Claude Sonnet 3.7 – often realize when they’re being evaluated for alignment. Here’s an example of Claude's reasoning during a sandbagging evaluation, where it learns from documentation that it will not be deployed if it does well on a biology test:

37 181 1K 225K 539

Download Image

Daniel Litt @littmath

6 months ago

In this thread I want to share some thoughts about the FrontierMath benchmark, on which, according to OpenAI, some frontier models are scoring ~20%. This is benchmark consisting of difficult math problems with numerical answers. What does it measure, and what doesn't it measure?

26 130 935 257K 517

Download Image

USGS Volcanoes🌋 @USGSVolcanoes

6 months ago

Mesmerizing video from #Kilauea's episode 12. Both north & south vents were active when the video was taken on Mar 5. Lava fountains jetted to 300 ft (90 m). Episode 12 ended later that day but given current rates of reinflation, a new episode is expected in the next 4-6 days.

9 235 941 33K 38

Download Video

USGS Volcanoes🌋 @USGSVolcanoes

6 months ago

Lava from #Kilauea's episode 12, which ended on March 5, covered about two-thirds of Halemaʻumaʻu. Ooze-outs along the margin of the flow decreased yesterday & are inactive today (Mar 7). New episode possible in next 4-6 days.

5 73 343 18K 11

Download Video

Patrick McKenzie @patio11

6 months ago

Matt Levine on a central and somewhat intractable challenge in Comms, Legal, PR, and Compliance, which is exacerbated by the increasing difficulty of keeping secrets in the modern world:

10 31 563 45K 89

Download Image

Andrej Karpathy @karpathy

9 months ago

People have too inflated sense of what it means to "ask an AI" about something. The AI are language models trained basically by imitation on data from human labelers. Instead of the mysticism of "asking an AI", think of it more as "asking the average data labeler" on the…

561 2K 14K 1.8M 5K

Misha Teplitskiy | Science of Science @MishaTeplitskiy

10 months ago

People usually think replication attempts in science are rare. Journals don't publish replications, so scientists don't do them. In reality there are countless replication attempts (and failures), it's just PhD students assume they did something wrong journals.plos.org/plosone/articl…

32 310 2K 433K 438

Download Image

Science Banana @literalbanana

10 months ago

A detailed review of the "nutrients in crops are declining" issue that's pretty interesting and not just a sketchy chart sciencedirect.com/science/articl…

5 32 277 26K 158

Download Image

Dr. Alexander S. Burns @KKriegeBlog

10 months ago

Fictional Data Scientists: With the power of psychohistory, I will accurately predict the collapse and rebirth of galactic civilization. Real Data Scientists: We've had hundreds of pollsters and every AI we got working on this election and the best we can say is "real close."