josh :) (@joshycodes) - Twitter Profili | Zamantika Mersobahis Locabet

Sabitlenmiş Tweet

josh :)@joshycodes·10 Mar

Can language models tell when their activations have been modified? Several recent papers found dim traces of this in frontier models. We asked: how easy is it to make this capability robust? Alarmingly easy. New paper with @DavidDAfrica 🧵

English

1

0

18

1.2K

josh :)@joshycodes·1d

Who in the world is making these nonsensical benchmark graphs… why in the world is higher cost on the left?!? @cursor_ai

English

0

5

199

josh :)@joshycodes·2d

we got Google Translate for Larry David before gta 6

Nick Khami@skeptrune

we got google translate for linkedin before gta 6

English

0

4

177

josh :)@joshycodes·2d

@askalphaxiv Please, make a cli instead

English

0

203

alphaXiv@askalphaxiv·3d

Introducing MCP for arXiv Let your research agents stand on the shoulders of giants Fast multi-turn retrieval, keyword search, and embedding search tools across millions of arXiv papers 🚀

English

69

388

3K

238.2K

josh :)@joshycodes·2d

midjourney, eternally damned to create images from the prompts of others —weird 1000

Midjourney@midjourney

Today we're starting to test an early version of our V8 model with our community. It's much better at following prompts, 5x faster, has native 2K modes, improved text rendering and the best personalization, sref, and moodboard performance ever. Have fun!

English

0

2

79

josh :)@joshycodes·5d

@biosemiote it worked perfectly!

English

0

3

39

Biosemiote@biosemiote·5d

@joshycodes godspeed

English

1

0

2

62

josh :)@joshycodes·5d

Tonight I’m trying the world’s first genetically engineered probiotic. It breaks down acetaldehyde, an unwanted byproduct of alcohol, in your gut If it works, I’ll have no hangover For science, I shall drink heavily and report how I feel in the morning

English

2

1

13

1.1K

josh :)@joshycodes·5d

@theGreenCoding we made it

English

0

8

Green Code@theGreenCoding·5d

@joshycodes We need an update on this, please!

English

1

0

1

30

josh :)@joshycodes·5d

@carobunga @DavidSKrueger x.com/joshycodes/sta…

josh :)@joshycodes

If Bilbo asked Grok

QME

0

41

5.4K

toe, lover@carobunga·5d

ZXX

46

4.1K

39.3K

919.7K

josh :)@joshycodes·5d

As for my sleep, not great, but not bad either!!!

English

0

66

josh :)@joshycodes·5d

After 5 shots, 1/3rd of sake, and 5 beers… I feel perfectly fine this morning! In fact, I’m about to go play basketball! zbiotics work!

josh :)@joshycodes

Tonight I’m trying the world’s first genetically engineered probiotic. It breaks down acetaldehyde, an unwanted byproduct of alcohol, in your gut If it works, I’ll have no hangover For science, I shall drink heavily and report how I feel in the morning

English

4

0

10

362

josh :)@joshycodes·5d

@steveruizok I believe it’s example A because you won’t have the overhead of dependency tracking. However, I think this would be negligible at scale

English

0

706

Steve Ruiz@steveruizok·5d

genuine question for react heads, which of these is most performant? Consider that we may have thousands of them all receiving new x y props on each frame.

English

48

6

316

74.5K

josh :)@joshycodes·5d

@allTheYud Epistemic status?

Română

1

0

4

594

Eliezer Yudkowsky@allTheYud·5d

Anthropic-DoW negotiations deteriorate further as Amodei retorts to Hegseth, "yo momma's ass be a supply risk, too much of it be gettin' supplied"

English

8

4

115

6.4K

josh :)@joshycodes·6d

@Plinz @rawmilkhoney

QME

0

1

90

Joscha Bach@Plinz·6d

@rawmilkhoney those who don’t think they have a 130+ iq are silent, those who actually have a 130+ iq are too smart to say something

English

19

1

157

7.4K

♡ Honey ♡@rawmilkhoney·13 Mar

Why is everyone on Twitter convinced they have a 130+ IQ

English

770

94

1.9K

194.6K

josh :)@joshycodes·12 Mar

@emil44hr May not be a bad start :x lesswrong.com/posts/38RcAQez…

English

0

338

emil@emil44hr·11 Mar

Sharbel@sharbel

I’m building an autoresearcher to test every single trading strategy possible and find the best one possible. This is a massive unlock for predictive analysis. Wish me luck. 🤝

ZXX

25

70

2.2K

129.7K

josh :)@joshycodes·12 Mar

@trashh_dev It’s all them shonens

English

0

19

trash@trashh_dev·12 Mar

lol everyone on this thing thinks they’re the main character

English

73

9

262

11.1K

josh :)@joshycodes·11 Mar

I ended up preferring human writing because I can’t stand the “not X. Y”

Kevin Roose@kevinroose

We made a blind taste test to see whether NYT readers prefer human writing or AI writing. 86,000 people have taken it so far, and the results are fascinating. Overall, 54% of quiz-takers prefer AI. A real moment! nytimes.com/interactive/20…

English

0

3

172

josh :)@joshycodes·10 Mar

Also, @LedermanHarvey: x.com/LedermanHarvey…

Harvey Lederman@LedermanHarvey

Can large language models *introspect*? In a new paper, @kmahowald and I study the MECHANISM of introspection in big open-source models. tldr: Models detect internal anomalies through DIRECT ACCESS, but don't know what the anomalies are. And they love to guess “apple” 🍎

English

0

3

85

josh :)@joshycodes·10 Mar

Other amazing and similar work by @voooooogel: x.com/voooooogel/sta…

thebes@voooooogel

we (@acsresearchorg) expanded this into a larger paper! (and my first proper paper.) we added some new experiments and found an interesting correlation. arxiv.org/abs/2602.20031

English

1

0

3

109

josh :)@joshycodes·10 Mar

Can language models tell when their activations have been modified? Several recent papers found dim traces of this in frontier models. We asked: how easy is it to make this capability robust? Alarmingly easy. New paper with @DavidDAfrica 🧵

English

1

0

18

1.2K

josh :)

Keşfet