Hacker News: rrhjm53270

New comment by rrhjm53270 in "What political censorship looks like inside an LLM's weights (Qwen 3.5)"

rrhjm53270 — Tue, 19 May 2026 02:21:23 +0000

Ironic

New comment by rrhjm53270 in "Iran will impose fees on subsea internet cables in Strait of Hormuz"

rrhjm53270 — Mon, 18 May 2026 21:54:07 +0000

Putin and Trump have already made the world chaotic enough; don't let Xi get involved too.

New comment by rrhjm53270 in "The local LLM ecosystem doesn’t need Ollama"

rrhjm53270 — Thu, 16 Apr 2026 10:15:23 +0000

It is a bit off-topic, but would it possible to provide a light mode for this blog? I used to work during the day time, and my pupils had to contract to read, making it a very poor reading experience.

New comment by rrhjm53270 in "Online astroturfing: A problem beyond disinformation (2022)"

rrhjm53270 — Sat, 14 Mar 2026 21:11:10 +0000

As an individual, this paper suggests that I must move beyond "Is this news true?" to "Why does it look like everyone agrees with this?"

New comment by rrhjm53270 in "There is an AI code review bubble"

rrhjm53270 — Mon, 26 Jan 2026 20:14:23 +0000

Why not let AI write the code and then have it reviewed by humans? If you use AI to review my code, then you can't stop me from using another AI to refute it: this only foreshadows the beginning of internal friction.

New comment by rrhjm53270 in "How well do you know C++ auto type deduction?"

rrhjm53270 — Mon, 15 Dec 2025 21:43:31 +0000

A trick that I used was

``` struct dummy{}; dummy d = ANYTHING_YOU_WANT_GO_GET_THE_TYPE; ```

Compile it with g++, and get the type info from compilation error -:)

New comment by rrhjm53270 in "Google's Genie is more impressive than GPT5"

rrhjm53270 — Fri, 08 Aug 2025 20:49:27 +0000

It is very interesting that among top-20 models, all non-proprietary ones are from China.

New comment by rrhjm53270 in "Show HN: I built a tensor library from scratch in C++/CUDA"

rrhjm53270 — Wed, 18 Jun 2025 19:27:58 +0000

Do you have any plan for the serialization and deserialization of your tensor and nn library?

Perplexity post-trained deepseek-r1 to remove Chinese Communist Party censorship

rrhjm53270 — Wed, 19 Feb 2025 21:01:55 +0000

Article URL: https://huggingface.co/perplexity-ai/r1-1776

Comments URL: https://news.ycombinator.com/item?id=43107627

Points: 5

# Comments: 1

New comment by rrhjm53270 in "LaTeX.css – Make your website look like a LaTeX document"

rrhjm53270 — Mon, 02 Dec 2024 12:13:59 +0000

I prefer write LaTex document, compile to pdf and then convert pdf to html using pdf2htmlEX.

New comment by rrhjm53270 in "A not so fast implementation of cosine similarity in C++ and SIMD"

rrhjm53270 — Wed, 16 Oct 2024 20:09:22 +0000

Thank you for sharing such an interesting work.

A little comment: adding some more aggressive optimization optimization options to simd C++ code to see the performance difference.

On my side with a AMD Ryzen 9 7900X3D CPU, I have

- 0.0592569 ms for `-O3 -march=native` option, and - 1.7741e-05 ms for `-funsafe-math-optimizations -Ofast -flto=auto -pipe -march=native`

New comment by rrhjm53270 in "Run 70B LLM Inference on a Single 4GB GPU with This New Technique"

rrhjm53270 — Mon, 04 Dec 2023 09:03:34 +0000

I did roughly the same thing in one of my hobby project https://github.com/fengwang/nnl. But in stead of using SSD, I load all the weights to the host memory, and while inferencing the model layer by layer, I asynchronously copy memory from global to shared memory in the hope of better performance. However, my approach is bounded by the PCI-E bandwidth.