<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Hacker News: richiejp</title><link>https://news.ycombinator.com/user?id=richiejp</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Wed, 05 Aug 2026 01:10:03 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=richiejp" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by richiejp in "Why we write our own C and C++ inference engines"]]></title><description><![CDATA[
<p>Getting any performance improvement over vLLM is just a bonus although the RAM reductions are very nice. The main point is that we find deploying Python based projects to be a pain. I can't imagine how you know what files are unreachable in a dynamic language, but in any case that doesn't begin to solve issues like PyTorch requiring a C++ compiler at runtime to generate optimised CPU code. This isn't exactly embedded device friendly.<p>Having said that if you know of some magic bullet to fix the problems with the Python ecosystem (that vLLM and Pytorch are apparently unaware of) then I will be happy to try it because our CI is constantly blocked by Python related issues.</p>
]]></description><pubDate>Mon, 03 Aug 2026 19:42:36 +0000</pubDate><link>https://news.ycombinator.com/item?id=49160474</link><dc:creator>richiejp</dc:creator><comments>https://news.ycombinator.com/item?id=49160474</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49160474</guid></item><item><title><![CDATA[New comment by richiejp in "Why we write our own C and C++ inference engines"]]></title><description><![CDATA[
<p>I can only assume you are talking about LocalAI (not vllm.cpp) which is modular and contains adapters for many different AI engines written in various languages. LocalAI's core is written in pure Go and you choose what engines to install at runtime.<p>vLLM.cpp's runtime is pure C++, but AOT Triton compilation is done at build time in Python.</p>
]]></description><pubDate>Mon, 03 Aug 2026 18:48:34 +0000</pubDate><link>https://news.ycombinator.com/item?id=49159857</link><dc:creator>richiejp</dc:creator><comments>https://news.ycombinator.com/item?id=49159857</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49159857</guid></item><item><title><![CDATA[C++ Vs Rust: Which is better for writing AI/ML code with LLMs]]></title><description><![CDATA[
<p>Article URL: <a href="https://richiejp.com/rust-vs-cpp-llm-coded-model-conversion-from-pytorch">https://richiejp.com/rust-vs-cpp-llm-coded-model-conversion-from-pytorch</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=49062785">https://news.ycombinator.com/item?id=49062785</a></p>
<p>Points: 2</p>
<p># Comments: 0</p>
]]></description><pubDate>Sun, 26 Jul 2026 21:50:30 +0000</pubDate><link>https://richiejp.com/rust-vs-cpp-llm-coded-model-conversion-from-pytorch</link><dc:creator>richiejp</dc:creator><comments>https://news.ycombinator.com/item?id=49062785</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49062785</guid></item><item><title><![CDATA[Show HN: LocalVQE: Tiny ~1M param audio model that cancels echo and noise]]></title><description><![CDATA[
<p>Article URL: <a href="https://huggingface.co/spaces/LocalAI-io/LocalVQE-demo">https://huggingface.co/spaces/LocalAI-io/LocalVQE-demo</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47973955">https://news.ycombinator.com/item?id=47973955</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Fri, 01 May 2026 12:21:31 +0000</pubDate><link>https://huggingface.co/spaces/LocalAI-io/LocalVQE-demo</link><dc:creator>richiejp</dc:creator><comments>https://news.ycombinator.com/item?id=47973955</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47973955</guid></item><item><title><![CDATA[New comment by richiejp in "[dead]"]]></title><description><![CDATA[
<p><a href="https://github.com/mudler/LocalAGI">https://github.com/mudler/LocalAGI</a></p>
]]></description><pubDate>Fri, 02 May 2025 08:44:29 +0000</pubDate><link>https://news.ycombinator.com/item?id=43867521</link><dc:creator>richiejp</dc:creator><comments>https://news.ycombinator.com/item?id=43867521</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=43867521</guid></item><item><title><![CDATA[Show HN: WebGPU Bitbanging 1D Reversible Automata]]></title><description><![CDATA[
<p>Recently I posted on here my automata implementation: <a href="https://news.ycombinator.com/item?id=42401976">https://news.ycombinator.com/item?id=42401976</a><p>It seems like hackers liked it and I got some interesting comments, so I decided to finally convert it to something that can run in the browser.<p>After this I'm interested in higher dimensional automata, but I vaguely remember from reading Wolfram's book that adding more dimensions makes it harder to find interesting rules, although it is possible to scan generated results and measure the entropy to see if it looks interesting. But anyway does anyone have advice or pointers on 3D automata?</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=42613063">https://news.ycombinator.com/item?id=42613063</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Mon, 06 Jan 2025 17:55:23 +0000</pubDate><link>https://richiejp.com/1d-reversible-automata-webgpu</link><dc:creator>richiejp</dc:creator><comments>https://news.ycombinator.com/item?id=42613063</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=42613063</guid></item><item><title><![CDATA[New comment by richiejp in "Bitbanging 1D Reversible Automata"]]></title><description><![CDATA[
<p>Yes it does. Even discarding arguments that different rules in the same class are the same automaton and multiple cells are the same automaton, the article includes reversible and non-reversible automata which are distinct classes.</p>
]]></description><pubDate>Fri, 13 Dec 2024 06:51:08 +0000</pubDate><link>https://news.ycombinator.com/item?id=42406454</link><dc:creator>richiejp</dc:creator><comments>https://news.ycombinator.com/item?id=42406454</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=42406454</guid></item><item><title><![CDATA[New comment by richiejp in "Bitbanging 1D Reversible Automata"]]></title><description><![CDATA[
<p>could you provide a link to the code?</p>
]]></description><pubDate>Thu, 12 Dec 2024 21:30:10 +0000</pubDate><link>https://news.ycombinator.com/item?id=42403559</link><dc:creator>richiejp</dc:creator><comments>https://news.ycombinator.com/item?id=42403559</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=42403559</guid></item><item><title><![CDATA[Bitbanging 1D Reversible Automata]]></title><description><![CDATA[
<p>Article URL: <a href="https://richiejp.com/1d-reversible-automata">https://richiejp.com/1d-reversible-automata</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=42401976">https://news.ycombinator.com/item?id=42401976</a></p>
<p>Points: 97</p>
<p># Comments: 22</p>
]]></description><pubDate>Thu, 12 Dec 2024 18:41:24 +0000</pubDate><link>https://richiejp.com/1d-reversible-automata</link><dc:creator>richiejp</dc:creator><comments>https://news.ycombinator.com/item?id=42401976</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=42401976</guid></item><item><title><![CDATA[New comment by richiejp in "It's easier to 10x code performance than 2x it"]]></title><description><![CDATA[
<p>It's definitely under rated that data locality may matter more than time or space complexity.<p>Most software though is doing really dumb stuff on I/O, or just doing work that doesn't need to be done.</p>
]]></description><pubDate>Sat, 07 Dec 2024 16:54:58 +0000</pubDate><link>https://news.ycombinator.com/item?id=42350980</link><dc:creator>richiejp</dc:creator><comments>https://news.ycombinator.com/item?id=42350980</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=42350980</guid></item><item><title><![CDATA[It's easier to 10x code performance than 2x it]]></title><description><![CDATA[
<p>Article URL: <a href="https://richiejp.com/how-to-10x-most-software">https://richiejp.com/how-to-10x-most-software</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=42350437">https://news.ycombinator.com/item?id=42350437</a></p>
<p>Points: 2</p>
<p># Comments: 2</p>
]]></description><pubDate>Sat, 07 Dec 2024 15:38:25 +0000</pubDate><link>https://richiejp.com/how-to-10x-most-software</link><dc:creator>richiejp</dc:creator><comments>https://news.ycombinator.com/item?id=42350437</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=42350437</guid></item><item><title><![CDATA[Fuzzy Sync: Winning a rare data race]]></title><description><![CDATA[
<p>Article URL: <a href="https://richiejp.com/a-rare-data-race">https://richiejp.com/a-rare-data-race</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=24598275">https://news.ycombinator.com/item?id=24598275</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Sat, 26 Sep 2020 12:24:09 +0000</pubDate><link>https://richiejp.com/a-rare-data-race</link><dc:creator>richiejp</dc:creator><comments>https://news.ycombinator.com/item?id=24598275</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=24598275</guid></item></channel></rss>