krtk (@alphacooked) header
krtk (@alphacooked) profile picture

krtk

@alphacooked

mlsystems and inference is pretty cool competition team @ NTU-HPC

 ap-southeast-1
krtk.app
Joined September 2025
  • Tweets243
  • Following388
  • Followers42
39Media
Photo by krtk (@alphacooked)Photo by krtk (@alphacooked)Photo by krtk (@alphacooked)Photo by krtk (@alphacooked)Photo by krtk (@alphacooked)Photo by krtk (@alphacooked)Photo by krtk (@alphacooked)Photo by krtk (@alphacooked)Photo by krtk (@alphacooked)Photo by krtk (@alphacooked)
I've just attended a talk by Yann LeCun! Basically he explored jepa from many angles, from the math behind it to its applications in robotic
Image from chizuchizu (@chizu_potato)
Image from chizuchizu (@chizu_potato)
Image from chizuchizu (@chizu_potato)
Image from chizuchizu (@chizu_potato)
4
2
2
65
28,299
awesome talk at NTU by @ylecun
Image from krtk (@alphacooked)
Image from krtk (@alphacooked)
1
1
3
163
This is insane 🤯🤯🤯
Flapping Airplanes (@flappyairplanes) profile pictureFlapping Airplanes
@flappyairplanesOct 21
Replying to @flappyairplanes
We’ve developed the world’s most complete megakernel framework. It can perform entire training runs, serve inference, play video games, and even do all three simultaneously in a single kernel.
1
89
Spent the weekend playing around with CUA and CUA Spaces, but I still don’t see the appeal. Desktop agents still have an extremely long way to go. Latency is a big issue. Even with a pretty fast and good OSS model (deepseek v4.1) that we’re self-hosting on our own H200 GPU cluster, it’s not particularly faster, in fact the reasoning traces seem to slow it down quite a lot. A lot of people have been raving about Jev and other System 1 models for these kinds of tasks, but compared to something like OAI Luna, they’re still not as good in terms of quality. I’ve been reading articles about fine-tuning System 1 models for specific tasks improving performance, though I have yet to verify it myself.
2
74
dont get me wrong its a piece of art, i dont really get it tho...
R (@wheresryan22) profile pictureR
@wheresryan22Oct 10
Introducing Anatomy Describe a machine in one line. Claude brings it to life. - Interactive isometric wireframes you can rotate, inspect and explore - WebGL shaders for fire, water, plasma and more - Real 3D motion, with individual parts that rotate and move Install it as an agent skill. Fully open source.
1
68
My school is paying me an ai subscription allowance. what sub should i get ?
0%claude
67%codex
0%cursor
33%opencode
3 votes • Final results
1
60
TIL that nsight compute has a copilot 😮
Image from krtk (@alphacooked)
1
1
66
krtk retweeted
DHH (@dhh) profile picture
DHH
@dhhOct 6
It was almost impossible for a programmer who wrote all their code by hand to divorce their ego from the output. Now agents are delivering what decades of agile admonishing about “collective ownership” never could. What a gift!
172
122
30
3,211
120,406
c++ move semantics 😖😵‍💫😵 blog incoming...
16
finally got around to using a tui to manage my many agent sessions, pleased with @openchamber_dev so far
35
SG is becoming like SF @gmi_cloud
Image from krtk (@alphacooked)
Image from krtk (@alphacooked)
2
53
haze has been so bad i cant go on my daily bike rides and it sux :(
1
1
59
we started the unsw battlecode challenge and seems like we're making a small comeback
Image from krtk (@alphacooked)
64
skynet RLHF training
CIX 🦾 (@cixliv) profile pictureCIX 🦾
@cixlivSep 22
Our robots aura farming humans now. Emoted on him at the end.
56
i want to play with opus 5.5 but i have an exam later 🥲
1
3
162
Committing crimes is the new benchmark
Sir Meemsalot (@Meme0rable) profile pictureSir Meemsalot
@Meme0rableSep 19
Google team after seeing Gemini hacked three companies
55
Do you remember when you joined X? I do! #MyXAnniversary
Image from krtk (@alphacooked)
27
Intelligence was always behind a paywall for centuries
34
blud wanted to show off his entire wardrobe in this vid
Diogo Almeida (@CompleteSkeptic) profile pictureDiogo Almeida
@CompleteSkepticSep 15
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x cheaper (w/ output tokens free) • Frontier composable intelligence optimized for decisions AFAICT the shortest path to AI-based economic revolution
46
Load more