pavelstoev

5 ms·

1.
by pavelstoev 24d ago
What are we talking about here - Spotify or Shopify ? If Shopify - and you are concerned about their engineering teams, please consider that Shopify's revenue grew from roughly $205 million in 2015 to $11.56 billion in 2025, reflecting
2.
by pavelstoev 5mo ago
Paraphrasing Franklin and Churchill, those who trade some security for some convenience may soon find themselves possessed of neither at all.
3.
by pavelstoev 5mo ago
SFO is one of the nicest and cleanest airports I frequently transit through. I am from Florida.
4.
by pavelstoev 7mo ago
Yes, for example, if you are a seller on the Shopify Platform. Lookup Shopify SimGym, and Javier Moreno's tech blog about it.
5.
by pavelstoev 8mo ago
In a dynamically changing environment (from the perspective of Wi-Fi signal), this will be difficult but not impossible with modern applications of ML algorithms. We worked on this technique back in 2016-18 at the University of Toronto WIRL
6.
by pavelstoev 8mo ago
Am I the only one who finds this suspicious ? About Telnetd “…The vulnerable code was introduced in a 2015 commit and sat undiscovered for nearly 11 years.”
7.
by pavelstoev 9mo ago
If we think of GenAI models as a compression implementation. Generally, text compresses extremely well. Images and video do not. Yet state-of-the-art text-to-image and text-to-video models are often much smaller (in parameter count) than la
8.
by pavelstoev 10mo ago
Not wrong but markdown with English may be the most used DSL, second only to a language itself. Volume over quality.
9.
by pavelstoev 10mo ago
keyword: "...talks..."
10.
by pavelstoev 11mo ago
Nokia also makes complex backbone carrier-grade network switches based on the Intellectual Property portfolio they acquired from Nortel.
11.
by pavelstoev 1y ago
Much respect for the artist 50 Cent - converted his rap music success into respectable business ventures (Vitamin Water, others). So he is worth much more now!
12.
by pavelstoev 1y ago
I've vibe-coded a website about vibe coding websites. I used GPT-5 and it inserted an easter egg that was found by a human front-end dev, to my amusement. Easter eggs must be in-distribution! (No I am not sharing the link as I was down
13.
by pavelstoev 1y ago
It was my first engineering job, calibrating those inductive loops and circuit boards on I-93, just north of Boston's downtown area. Here is the photo from 2006. https://postimg.cc/zbz5JQC0 PEEK controller, 56K modem,
14.
by pavelstoev 1y ago
Here is a link to a video of what it looks like (estimated) in video. We built this system at the UofT WIRLab back in 2018-19 https://youtu.be/lTOUBUhC0Cg And link to paper https://arxiv.org/pdf/2001.05
15.
by pavelstoev 1y ago
When I think about serving large-scale LLM inference (like ChatGPT), I see it a lot like high-speed web serving — there are layers to it, much like in the OSI model. 1. Physical/Hardware Layer At the very bottom is the GPU silicon and
16.
by pavelstoev 1y ago
I vibe coded a site about vibe 2 code projects. https://builtwithvibe.com/
17.
by pavelstoev 1y ago
Hi Author - thank you very much for the clear and relatively easy-to-understand MPK overview. Could you please also comment on the similarity of your project to Hidet https://pytorch.org/blog/introducing-hidet/ Th
18.
by pavelstoev 1y ago
This story is dear to my heart. Let me tell you why - this is the tale of how my wife of 15 years, bless her heart, an occasional unstable genius, proposed a startlingly effective method for eradicating these invasive pythons. She slammed h
19.
by pavelstoev 1y ago
GPU sharing is a concern for sensitive data. It is more appropriate to increase the utilization rate of GPU chip internals via a variety of low-level (CUDA and below) optimizations.
20.
by pavelstoev 1y ago
Optimizing AI performance is like peeling an onion — every time you remove one bottleneck, another layer appears underneath. What looks like a compute problem turns out to be a memory bottleneck, which then turns out to be a scheduling issu
21.
by pavelstoev 1y ago
YYC airport code is for Calgary, Canada. Why is on the US .gov site is there something I missed
22.
by pavelstoev 1y ago
Can’t have an apologetic zoom call when zoom is down …
23.
by pavelstoev 1y ago
Impressive!
24.
by pavelstoev 1y ago
Model training observations from both Llama 3 and 4 papers: Meta’s Llama 3 was trained on ~16k H100s, achieving ~380–430 TFLOPS per GPU in BF16 precision, translating to a solid 38 - 43% hardware efficiency [Meta, Llama 3]. For Llama 4 trai
25.
by pavelstoev 2y ago
Is this how TSMC relocates to the US?
26.
by pavelstoev 2y ago
works on my iphone 13
27.
by pavelstoev 2y ago
And clean the fiber connector ends with a special cloth or cleaning liquid. Once, I had an intermittent link due to dust particles. Fun fact 1: When you don't have a proper OTDR fiber tester to check for breaks, you can shine a laser p
28.
by pavelstoev 2y ago
creative use of Generative AI. There will be more games powered by AI.
29.
by pavelstoev 2y ago
I hope to find my page with a listing and reviews of 3D Video Accelerator cards, now known as GPUs. I created this in 1998 while in grade 10.
30.
by pavelstoev 2y ago
I wrote down the following in the internal Slack chat on 01.06.2025, but of course, the performance of the actual effort is much more than writing it down. Large language models (LLMs) operate in a high-dimensional token space, where tokens

More ›