VioP (@AcousimHss) header
VioP (@AcousimHss) profile picture

VioP

@AcousimHss

the more you laugh the more you cry the more you cry the more you laugh bewarse | prev @rumik_ai

silverjinx.bearblog.dev/
Joined January 2025
  • Tweets4.4K
  • Following1.2K
  • Followers228
539Media
Photo by VioP (@AcousimHss)Photo by VioP (@AcousimHss)Photo by VioP (@AcousimHss)Photo by VioP (@AcousimHss)Photo by VioP (@AcousimHss)Photo by VioP (@AcousimHss)Photo by VioP (@AcousimHss)Photo by VioP (@AcousimHss)
This is peak work as I do slavery with claude while I watch DLNS
Ryan Morrissey (@ryan_morr) profile pictureRyan Morrissey
@ryan_morr15h
my cto is watching league and managing five agents at once we may be GMI
Image from Ryan Morrissey (@ryan_morr)
2
84
Bro my niece is blasting the crackers with a lighter???? I mean I get the efficiency but half the fun was with agarbatti taking time to light it up and throw
3
50
Happy diwali!!!!!
2
31
> DeepSeek bypassed sandbox network restrictions to call OpenRouter surprised openai didn't escape the sandbox this time
Image from VioP (@AcousimHss)
Justin Lo (@loresjpg) profile pictureJustin Lo
@loresjpgNov 7
Nice DeepSWE score! Does it survive a lowercase filename? Introducing MaxxBench: 89 DeepSWE tasks with small changes to the instructions or starting code. Same engineering problem, slightly different requirement. Models clustered around 74% on DeepSWE end up 23 pp apart. Good at coding, or good at DeepSWE? 🧵
Image from Justin Lo (@loresjpg)
2
77
someone do this for deadlock
Ado (@adocomplete) profile pictureAdo
@adocompleteNov 6
WoW Forever with a custom UI add on to show me what my Claude agents are doing and ability to kick off new tasks from within the game. 👀 All work and all play.
Image from Ado (@adocomplete)
2
81
This paige shit is easyyy Get me in the leagues #needmeaduo
Image from VioP (@AcousimHss)
54
new benchmark for my idiotic ranked players : if they cant pass even one badge , take away their internet access now!!!!
Image from VioP (@AcousimHss)
Anthropic (@AnthropicAI) profile pictureAnthropic
@AnthropicAINov 6
Memory is what allows Claude to learn from experience. In a new Engineering Blog post, we share how “dreaming” helps Claude organize what it’s learned, correct its mistakes, and find what it needs more easily the next time. anthropic.com/research/compo…
The components of agent memory

The components of agent memory

For Claude to learn from experience, it has to decide what’s worth remembering before it knows what’s next. Here’s how “dreaming” helps it organize memory and correct mistakes.

anthropic.com
49
if you really think about it, a good employee isn’t the smartest one, it’s the one who knows where to find things and who to ask. here the who changes and so does the env!
Image from VioP (@AcousimHss)
Lorenzo Xiao (@lrzneedresearch) profile pictureLorenzo Xiao
@lrzneedresearchNov 7
Replying to @edwardjhu
algoroxyolo.github.io/blog/2…

Knowing How Is Not Knowing What

What RL environments need to supply: organizational context, domain rules and connected information.

algoroxyolo.github.io
2
108
Correct me if I'm wrong, model is really good for just 12T tokens of training Is it not Not to mention the NVFP4 training as well
Thinking Machines (@thinkymachines) profile pictureThinking Machines
@thinkymachinesNov 5
Introducing Inkling 2 Small. Scoring 31 on the AA Intelligence Index, it's a big jump over its predecessor. Inkling 2 Small is available for fine-tuning and inference on Tinker today, with full weight release planned soon. thinkingmachines.ai/news/ink…
Introducing Inkling 2 Small

Introducing Inkling 2 Small

Today we are introducing the second generation of Inkling models with the release of Inkling 2 Small.

thinkingmachines.ai
1
123
Wao everyone's getting the diwali gifts
35
Did anyone try DBC from polar bear???
45
VioP retweeted
fig (@figbrains) profile picture
fig
@figbrainsNov 5
We're releasing RealCUA, the first computer-use dataset of open-ended work in evolving environments. 600+ hours (and growing!) of real multi-objective work with interleaved tasks across thousands of steps in realistic desktops. 11-hour sample & access: data.fig.inc/realcua
21
41
8
429
55,853
Meet pylexnet : Neural net syntax highlighter for python containing 827 parameters all hand written by Fable and Opus which is quite accurate and fast!
Image from VioP (@AcousimHss)
1
1
8
206
more replies
HF : huggingface.co/Pranavz/pylex… github : github.com/PranavGovindu/pyl…
Pranavz/pylexnet · Hugging Face

Pranavz/pylexnet · Hugging Face

Machine learning model. Download, explore files and run it on Hugging Face.

huggingface.co
1
1
44
work inspired by @AlpinDale 's lexnet teapaw.org/AlpinDale/status/21177…
Alpin (@AlpinDale) profile pictureAlpin
@AlpinDaleNov 3
I had Fable 5.1 write a neural net to do syntax highlighting for C. It's about 984 parameters, and very fast! Fable wrote all the weights by "hand", of course. huggingface.co/alpindale/lex…
Image from Alpin (@AlpinDale)
22
playing deadlock anytime but late night is suicide
1
2
47
always start with evals. work from first principles
32
44
10
794
33,178
Load more