yu3zhou4

Thx, great read about Yemen from Maciej Cegłowski https://idlewords.com/2014/07/sana_a.htm

An open course on building high performance LLM inference engine! Hope to finish by the end of April

https://github.com/jmaczan/tiny-vllm

I love this article, the edge cases are where the seeming “simplicity” of floating-point numbers breaks down

Recently wrote a chapter in tiny-vllm course about floats in context of LLM inference, much shorter and not that deep as this one, for anyone interested in topic you might like it too https://github.com/jmaczan/tiny-vllm?tab=readme-ov-file#how-...

Honestly there is a lot for room of improvement in torch-webgpu for performance. Needs involvement of community but the opportunities are definitely there

Overall I think it’s a good perspective and worth reading to learn your perspective and experience. Just for anyone who doesn’t come from a good place, please be kind to yourself. You probably really need to work exceptionally hard and need to have top 1% resume. And articles like this can be quite depressing. Don’t let it you down

The author tries to be mindful about it and I appreciate it, but from a perspective of someone from a shithole it still might feel bad

Referring to:

> I've never gotten a job by applying to it. It's always been referrals or someone reaching out to me. So honestly, my resume is shit compared to my peers. I'm terrible at interviewing, and I've never done LeetCode. This is not a brag; it's just not my style. Am I a nepo baby? I don't know. Was I a morale hire? I'm pretty disagreeable. Was it merit? Also not sure

> I recognize that not everyone has access to the same networking opportunities, and the traditional job application process can be a valid and necessary path for many. But social media, when used correctly, is a great way to get an opportunity

Good for the maintainer, hope they find peace and do things just for their fun, without needing to deal with comments like that anymore

So in some sense we start recreating an operating system, or at least the userspace, within the Claude code. There was some name for this pattern but I can’t recall

Can confirm that it also happens in other complex systems! Still a lot of good time and the novelty factor helps with pushing through

Congrats and happy for you, you had a lot of fun and did something genuinely interesting

I’m recreating a tiny version of vLLM in C++ and CUDA from scratch (high throughput LLM inference server)

https://github.com/jmaczan/tiny-vllm

Liar paradox

I had to use AI to actually understand what you wrote it and I think it's an underrated comment

Thank you very much!

Please share prompts or md files so we can generate more!

I try my best to push ML things into WebGPU and I think it has a future, but performance is not there yet. I have little experience with Vulkan except toy projects, but WebGPU and Vulkan seem very similar

PyTorch compiler and runtime for WebGPU!

ngmi

At this point, I genuinely don't know what to learn next to not become obsolete when another Opus version gets released

This is hilarious, 5 years ago I built a very similar cli tool based on the same idea and with the same name (whosthere but in Polish, ktotu [0]). I wonder if you used AI to generate the project and the idea

Congrats for the execution, it looks more complete and feature rich and Go is a better choice for sure

[0] https://github.com/jmaczan/ktotu

Performance is not there yet, honestly. I haven’t focus on running things fast, but it’s one of next directs steps I’ll take. The problem with running it in a browser is that we need to find a way to run the PyTorch itself in a browser and to my best knowledge it’s not there yet. I’m looking at it to close the gap, feel welcome to reach out to collaborate if you’re interested!

PyTorch compiler and runtime for WebGPU https://github.com/jmaczan/torch-webgpu

On Safari mobile it shows a modal that can’t be scrolled nor closed

Good luck to everyone in achieving their goals and exploring new paths!

To me it's deep learning compilers since mid 2025. I am a person who can't learn just from reading books, so 80% of time I learn by doing (contribute to PyTorch) and 20% of time I read books (now: Engineering a Compiler from Keith Cooper and Linda Torczon) and talk to LLMs to fill gaps in my understanding.

My main quest now is to build a bridge [0] between PyTorch and universal GPU computing world - which I believe WebGPU might become. What it requires is to build is 1) a runtime for executing PyTorch ATen operations on WebGPU by running WGSL shaders and 2) a compiler, so you can use full PyTorch power with @torch.compile

[0] - https://github.com/jmaczan/torch-webgpu

WebGPU backend for PyTorch

https://github.com/jmaczan/torch-webgpu

Glad you enjoyed it, your comment made my day

Thanks. I'd like to learn the most important things in traditional compilers that are still relevant and translate these skills into ML compilers

Thanks!

Are you sure it’s an extinct art though? LLVM is flourishing, many interesting IRs come to life like MLIR, many ML-adjacent projects build their own compilers (PyTorch, Mojo, tinygrad), many big tech like Intel, AMD, Nvidia, Apple and others contribute to multiple different compilers, projects integrate one to another at different levels of abstraction (PyTorch -> Triton -> CUDA) - there is a lot of compilation going on from one language to another

Not to mention many languages in a mainstream that weren’t that popular 10 years ago - think Rust, Zig, Go

Is Dragon book still relevant? Do you recommend any other learning resources other than reading the source and contributing to llvm?

There a free online accredited master's degree in Financial Engineering (https://www.wqu.edu/mscfe), probably not for OP but maybe some find it useful for them

Are there any alternative mobile OSes actively developed? I remember Ubuntu Touch was the thing and something from Firefox, but not sure if they are continued?