Bing Xu (@bingxu

Bing Xu

699 posts

Bing Xu

@bingxu_

Founder & CEO @int21_ai. Founder & CEO @hippoml_com (acq'ed by NVIDIA). Built AITemplate, MXNet, CXXNet. Named GAN. Tweets are my own.

antinucleon.github.io

Joined October 2019

Pinned
Bing Xu
@bingxu_
Jun 16
I started INT21 two months ago, and I’m proud to announce that we’re coming out of stealth today with our first product: PTX Kernel Factory. We’ve created the first self-improving agent swarm for producing expert-level PTX kernels autonomously and at massive scale.
INT21
@int21_ai
Jun 16
Article
Introducing INT21 and PTX Kernel Factory
We are building self-improving AI systems for the software beneath modern AI. Our first product generates and optimizes low-level GPU software, then proves its work with tests and benchmarks. Today,...
INT21 | Self-Improving Compute Infrastructure
From int21.ai
14K
Bing Xu
@bingxu_
Jul 10, 2025
Grok 4 is smart. Tested a block sparse attention problem which both o3 and Claude opus failed. Grok 4 knows it.
2.1M
Bing Xu
@bingxu_
Feb 22, 2025
Just received a certificate from David. Unbelievable—I’ve been babysitting at home for 5 years. I hope to meet more friends in person once they’ve grown up.
57K
Bing Xu
@bingxu_
Aug 24, 2025
Cursor annual subscription is ending, and I canceled it. I don’t use public coding agents for a while, but thank you, Cursor. Back in 2019 summer bbq, @ilyasut told me software engineering jobs might be gone in our lifetime. I was fool didn’t believe it.
67K
Bing Xu
@bingxu_
Aug 23, 2025
Yesterday once more. I was the first people to enable MacBook GPU training, getting it to run at about one-quarter the speed of a P100 in 2016–2017 for fine-tuning models. After that, it was all mediocre politics with no real technical vision. I developed PTSD and took half a
NIK
@ns123abc
Aug 23, 2025
Apple was so early into the machine learning. they literally invented Siri 10+ years ago lmao what went wrong?
Readers added context
Apple didn’t „invent“ Siri. Siri was an already existing software that was acquired by Apple in 2010. en.m.wikipedia.org/wiki/Siri
155K
Bing Xu
@bingxu_
Feb 20, 2025
Replying to @SakanaAILabs
I quickly take a look of their report on phone, there are a few misleading parts: 1. Torch C++ code is not CUDA kernel, it is calling CUDNN under hood. 2. The highlighted example Conv3D GroupNorm, conv code is not generated at all. The speedup doesn’t make sense if numerical
36K
Bing Xu
@bingxu_
Feb 11, 2025
Replying to @mtrainier2020
Driving video...
75K
Bing Xu
@bingxu_
Feb 28, 2025
Shame time for US Tech companies again.
DeepSeek
@deepseek_ai
Feb 28, 2025
🚀 Day 5 of #OpenSourceWeek: 3FS, Thruster for All DeepSeek Data Access Fire-Flyer File System (3FS) - a parallel file system that utilizes the full bandwidth of modern SSDs and RDMA networks. ⚡ 6.6 TiB/s aggregate read throughput in a 180-node cluster ⚡ 3.66 TiB/min
21K
Bing Xu
@bingxu_
Jan 2, 2024
Quote From Ian Buck, VP at NVIDIA: Today’s generative AI applications require not just speed, but also intelligent model optimization and efficient computing to deliver breakthrough experiences. HippoML’s approach future boosts the capabilities of the NVIDIA Tensor Core GPUs
HippoML
@hippoml_com
Jan 2, 2024
Introducing PrivateCanvas. Harness the power of your local GPU for contiguous editing and generating with cutting-edge models like Large Language Model, SDXL, Segment Anything, and GANs. Experience top-tier performance with minimal hardware demands. blog.hippoml.com/super-ai-creat…
143K
Bing Xu
@bingxu_
Aug 2, 2025
When AGI is achieved, ex Meta AI infra mates should organize a big party to get together again. Unbelievable intelligence density at its peak time.
8.7K
Bing Xu
@bingxu_
Feb 28, 2025
Replying to @mtrainier2020
半斤八两大加速时代。“通缩不好吗老百姓不都喜欢东西便宜”
10K
Bing Xu
@bingxu_
Feb 11, 2025
Replying to @mtrainier2020
That’s why I think no one else on earth can achieve self driving other than $TSLA
21K
Bing Xu
@bingxu_
Feb 20, 2025
Replying to @BlasianHokage and @SakanaAILabs
We published a blog last week:
Automating GPU Kernel Generation with DeepSeek-R1 and Inference Time Scaling | NVIDIA Technical Blog
From developer.nvidia.com
4.5K
Bing Xu
@bingxu_
Feb 12, 2025
NVIDIA AI Developer
@NVIDIAAIDev
Feb 12, 2025
Our engineers developed an AI coding agent using DeepSeek-R1 and inference-time scaling to automatically generate GPU attention kernels. Read how ➡️ nvda.ws/42V0mze
13K