tetsuo_cpp's profile picture. Compilers, infosec, systems-programming. Building compilers for ZK hardware. Senior Staff Supreme Engineer.

tetsuo.cpp

@tetsuo_cpp

Compilers, infosec, systems-programming. Building compilers for ZK hardware. Senior Staff Supreme Engineer.

置頂

Sometimes I like to put on my headphones and listen to electronic music so I feel like a leet hacker while I tell Claude to think hard and not make any mistakes.


Well, this is cursed.

Prediction: the next wave of programming languages will just be AI prompts. Someone's just released ai_bindgen (github.com/germangb/ai-bi…) into the Rust ecosystem, which allows annotation to create functionality with natural language:

timClicks's tweet image. Prediction: the next wave of programming languages will just be AI prompts.

Someone's just released ai_bindgen (github.com/germangb/ai-bi…) into the Rust ecosystem, which allows annotation to create functionality with natural language:


I think tile-based programming models are happening whether they're onboard with it or not. So the choice is really between letting something portable like Triton gain traction, or to win with their own solution where they're free to introduce NV specific primitives and throw…

Curious. Did Nvidia end the Cuda "moat" ? If they move to tiles like most other hardware, the AI kernels will be easier to port. x.com/nvidiaaidev/st…



tetsuo.cpp 已轉發

(reposts appreciated pls) hey, is anyone in the bay (preferably sf) willing/able to take care of a kitten for a little bit? while going to the ER today we took a stray in because he looks just like my kitty that passed away.. it made me cry and i felt really emotional about…


Very interesting! I wonder how this compares to Triton.

CUDA Tile has shipped! You can now `pip install cuda-tile`. I'm excited to see what y'all will build with it! Docs & resources: developer.nvidia.com/cuda/tile GitHub: github.com/NVIDIA/cutile-…



tetsuo.cpp 已轉發

call me anthropic the way i’m acquiring buns


A while ago, when I was working on databases, I was attracted to the idea of working on compression algorithms such as Zstd, Snappy, LZ4, etc. I think it's amazing that you can have such an outsized impact on the world with a (relatively) compact codebase.


Apparently, jemalloc is no longer being developed, and its GitHub repo is now archived. It's been in use for 20 years, that's a hell of a run! :-) jasone.github.io/2025/06/12/jem…


Do people play MUD (Multi User Dungeon) games anymore? I remember I used to play one called Achaea as a kid. They're essentially text-based MMORPGs.


Damn, too bad my irl stats aren't like this.

tetsuo_cpp's tweet image. Damn, too bad my irl stats aren't like this.

Wait am i intimidating??

CWood_sdf's tweet image. Wait am i intimidating??


I should start getting into engagement bait.


Seems like the hype is dying down, but I still can't wait to try Sora. I doubt the content will be any good, but I want to generate videos of my friends getting arrested, acting like assholes in public, etc. I feel like there's great potential for group chat banter.


I recently got time to look at this again and there are a lot of MLIR dialects on the GPU side of things! Afaict, my compilation pipeline is going to look something like this: - forth-to-mlir: translation from Forth source to my Forth MLIR dialect, the ops model Forth semantics…

Lately, I've been interested in learning about GPUs and some of the MLIR dialects used to target them. So I've started writing a Forth compiler for GPU kernels. I'll post updates as I go, this could be interesting. :-)



Finally got one! And yes, it is fast.

Damn, I've been trying to get my hands on a @cerebras Code subscription so I can get fast GLM 4.6 but they've been limiting plans due to capacity. There were meant to be more plans available today but they apparently sold out in 3 minutes!



What is happening right now… 😭

IF YOU CANT HIDE IT THEN YOU DONT OWN IT. This is the problem with crypto. ZERO KNOWLEDGE PROOF



The Compiler Company of Santa Clara ®

Time for the regular reminder that just as nvidia is really a compiler development company funded by selling chips, ARM is really a documentation company funded by selling IP.



What tokens are you more tired of hearing about?

LLM tokens %66.7
Shitcoins %33.3

6 票 · 最終結果


United States 趨勢

Loading...

Something went wrong.


Something went wrong.