<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: milansuk</title><link>https://news.ycombinator.com/user?id=milansuk</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Fri, 17 Apr 2026 19:16:45 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=milansuk" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by milansuk in "Go library for in-process vector search and embeddings with llama.cpp"]]></title><description><![CDATA[
<p>No need to use Ollama. LLama.cpp has its own OpenAI-compatible server[0] and it works great.<p>[0] <a href="https://github.com/ggerganov/llama.cpp#web-server">https://github.com/ggerganov/llama.cpp#web-server</a></p>
]]></description><pubDate>Wed, 30 Oct 2024 11:58:30 +0000</pubDate><link>https://news.ycombinator.com/item?id=41993932</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=41993932</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=41993932</guid></item><item><title><![CDATA[New comment by milansuk in "Gen AI Makes Legal Action Cheap – and Companies Need to Prepare"]]></title><description><![CDATA[
<p>Higher volume legal actions can only be successful as "peer-to-peer". If it goes through a courtroom, there is no way they can handle this kind of volume(even with AI tools). Imagine that the CEO is informed there are 20 new legal actions and the first court hearing will start 10-20 years from now when the CEO will not be part of the company anymore.</p>
]]></description><pubDate>Sat, 05 Oct 2024 16:30:40 +0000</pubDate><link>https://news.ycombinator.com/item?id=41751024</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=41751024</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=41751024</guid></item><item><title><![CDATA[New comment by milansuk in "Scramble: Open-Source Alternative to Grammarly"]]></title><description><![CDATA[
<p>> It doesn't do that fancy "make this text more professional"<p>I looked into the Scramble code[0] and it seems there are few pre-defined prompts(const DEFAULT_PROMPTS).<p>[0] <a href="https://github.com/zlwaterfield/scramble/blob/main/background.js">https://github.com/zlwaterfield/scramble/blob/main/backgroun...</a></p>
]]></description><pubDate>Wed, 18 Sep 2024 09:50:55 +0000</pubDate><link>https://news.ycombinator.com/item?id=41577749</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=41577749</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=41577749</guid></item><item><title><![CDATA[New comment by milansuk in "Breakthrough a step toward revealing hidden structure of prime numbers"]]></title><description><![CDATA[
<p>True. The only solution is to keep your data outside cloud(aka someone else's computer) no matter what encryption you use.</p>
]]></description><pubDate>Thu, 01 Aug 2024 10:50:30 +0000</pubDate><link>https://news.ycombinator.com/item?id=41127839</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=41127839</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=41127839</guid></item><item><title><![CDATA[New comment by milansuk in "Local First, Forever"]]></title><description><![CDATA[
<p>> But file syncing is a “dumb” protocol. You can’t “hook” into sync events, or update notifications, or conflict resolution. There isn’t much API; you just save files and they get synced. In case of conflict, best case, you get two files. Worst — you get only one :)<p>Sync services haven't evolved much. I guess, a service that would provide lower APIs and different data structures (CRDTs, etc.) would be a hacker's dream. Also, E2EE would be nice.<p>And if they closed the shop, I would have all the files on my devices.</p>
]]></description><pubDate>Tue, 25 Jun 2024 10:39:07 +0000</pubDate><link>https://news.ycombinator.com/item?id=40786755</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=40786755</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40786755</guid></item><item><title><![CDATA[New comment by milansuk in "Apple Intelligence for iPhone, iPad, and Mac"]]></title><description><![CDATA[
<p>This looks cool for v1! The only problem I see is most devices don't have much RAM, so local models are small and most requests will go to the servers.<p>Apple could use it to sell more devices - every new generation can have more RAM = more privacy. People will have real reason to buy a new phone more often.</p>
]]></description><pubDate>Mon, 10 Jun 2024 19:16:53 +0000</pubDate><link>https://news.ycombinator.com/item?id=40637425</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=40637425</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40637425</guid></item><item><title><![CDATA[New comment by milansuk in "Meta Llama 3"]]></title><description><![CDATA[
<p>I don't see any explanation for why they trained 8B instead of 7B.
I thought that If you have a 16GB GPU, you can put 14GB(7B*16bits) model into it, but how does it fit If the model is exactly 16GB?</p>
]]></description><pubDate>Thu, 18 Apr 2024 16:45:19 +0000</pubDate><link>https://news.ycombinator.com/item?id=40078146</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=40078146</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40078146</guid></item><item><title><![CDATA[New comment by milansuk in "Mixtral 8x22B"]]></title><description><![CDATA[
<p>The progress is insane. A few days ago I started being very impressed with LLM coding skills. I wanted Golang code, instead of Python, which you can see in many demos. The prompt was:<p>Write a Golang func, which accepts the path into a .gpx file and outputs a JSON string with points(x=tolal distance in km, y=elevation). Don't use any library.</p>
]]></description><pubDate>Wed, 17 Apr 2024 15:54:18 +0000</pubDate><link>https://news.ycombinator.com/item?id=40066434</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=40066434</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40066434</guid></item><item><title><![CDATA[New comment by milansuk in "Llm.c – LLM training in simple, pure C/CUDA"]]></title><description><![CDATA[
<p>This is an implementation of a transformer and in README it's presented as text->text. Tokens are just integers going in and out.<p>Is it possible to use it to train other types of LLMs(text->image, image->text, speech->text, etc.)?</p>
]]></description><pubDate>Tue, 09 Apr 2024 08:50:30 +0000</pubDate><link>https://news.ycombinator.com/item?id=39977466</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=39977466</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=39977466</guid></item><item><title><![CDATA[New comment by milansuk in "Gemma: New Open Models"]]></title><description><![CDATA[
<p>You can run Gemma and hundreds of other models(many fine-tuned) in llama.cpp. It's easy to swap to a different model.<p>It's important there are companies publishing models(running locally). If some stop and others are born, it's ok. The worst thing that could happen is having AI only in the cloud.</p>
]]></description><pubDate>Wed, 21 Feb 2024 14:07:25 +0000</pubDate><link>https://news.ycombinator.com/item?id=39453967</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=39453967</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=39453967</guid></item><item><title><![CDATA[New comment by milansuk in "Tesla Makes It Harder for Investors to Ignore Its Problems"]]></title><description><![CDATA[
<p>Agree. Spinning out the FSD and Optimus teams wouldn't accelerate progress on the new models. Also, they will make more money from humanoids than car robots.</p>
]]></description><pubDate>Thu, 25 Jan 2024 12:25:08 +0000</pubDate><link>https://news.ycombinator.com/item?id=39128791</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=39128791</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=39128791</guid></item><item><title><![CDATA[New comment by milansuk in "Yi-34B-Chat"]]></title><description><![CDATA[
<p>Yeah, they probably don't want their 'demo' links on the HN home page.</p>
]]></description><pubDate>Fri, 24 Nov 2023 18:12:27 +0000</pubDate><link>https://news.ycombinator.com/item?id=38406597</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=38406597</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=38406597</guid></item><item><title><![CDATA[New comment by milansuk in "Building a unikernel that runs WebAssembly – part 1"]]></title><description><![CDATA[
<p>> Believe it or not ...<p>I believe and agree with most of you wrote ;)<p>The main problem with HTML/CSS/JS is programmers want more than these languages offer. With WASM you can pick up language(must compile to .wasm) that fits your use case best. This is the freedom most programmers want.<p>There will always be programmers who will draw their custom buttons(instead of modifying DOM from WASM) and ignore accessibility. They can do this with JS as well, but most of them don't.</p>
]]></description><pubDate>Mon, 23 Oct 2023 17:11:09 +0000</pubDate><link>https://news.ycombinator.com/item?id=37988577</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=37988577</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=37988577</guid></item><item><title><![CDATA[New comment by milansuk in "Building a unikernel that runs WebAssembly – part 1"]]></title><description><![CDATA[
<p>Nice project! I love WASM. It's designed to be sandboxed and portable from day one. I wish WASM was invented instead of Javascript in the 90s. WASM will eat the world.<p>What I hope most is endurance. There are many programs that we are not able to run anymore. The best examples are probably older games. I hope WASM will change that, although I'm a little bit nervous about adding new features, because simple specs have a higher chance of surviving, but the future of binaries looks exciting.</p>
]]></description><pubDate>Mon, 23 Oct 2023 12:24:02 +0000</pubDate><link>https://news.ycombinator.com/item?id=37984653</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=37984653</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=37984653</guid></item><item><title><![CDATA[Show HN: Build Local-First Apps(WASM+SQLite)]]></title><description><![CDATA[
<p>Hey, everyone!<p>SkyAlt is a browser and SDK for Local-first applications at the top of SQLite files.<p>One and half months ago, I started working on the idea of rebuilding some software as services I use into local-first apps. SkyAlt is a combination of SQLite(storage), WASM(app binary) and JSON(settings, communication).<p>I decided to build my own 'browser', which has around ~10K LOC and has nothing to do with WebKit(HTML/CSS/JS) or Chromium. Apps use immediate mode GUI to draw and communicate with Skyalt. The layout is a flexible grid.
Apps are written in the Go language. Right now SDK is only for Go, but it should be simple to add more languages(which compile into WASM), when apis() settle.<p>I built a few example apps(7Gui, Map, Calendar, Table/DB editor). My approach is to build more apps across different use cases and add new features to current ones, which leads to better SDK, browser and developer experience.<p>For app development, you can use any IDE and debugger(or print to a terminal). When you make an app, you compile the app into binary which has a TCP socket into SkyAlt. For final shipping, you compile into WASM(portability & sandboxing) and the TCP socket is replaced with WASI(WASM IO). This accelerate development alot because You can use tools you're used to and can iterate quickly(go compiler is ~1sec vs tinygo(.wasm) is ~40sec).<p>You can find more information and 100% of code(Apache v2.0) in Github repo: https://github.com/MilanSuk/skyalt<p>What do you think? Let's talk about Local-first future!</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=37480897">https://news.ycombinator.com/item?id=37480897</a></p>
<p>Points: 7</p>
<p># Comments: 0</p>
]]></description><pubDate>Tue, 12 Sep 2023 13:29:30 +0000</pubDate><link>https://news.ycombinator.com/item?id=37480897</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=37480897</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=37480897</guid></item><item><title><![CDATA[New comment by milansuk in "The Age of AI has begun"]]></title><description><![CDATA[
<p>> Risks and problems with AI<p>One of the biggest risks right now is viruses created by transformers trained on known viruses and anti-virus binaries and databases.<p>Humans created relatively simple viruses which took down hospitals, etc.. Transformer can make new "super" viruses and keep fighting the anti-viruses/firewalls in real-time with high-quality custom attacks.</p>
]]></description><pubDate>Tue, 21 Mar 2023 20:09:23 +0000</pubDate><link>https://news.ycombinator.com/item?id=35251825</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=35251825</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=35251825</guid></item><item><title><![CDATA[New comment by milansuk in "Nostr: Notes and Other Stuff Transmitted by Relays"]]></title><description><![CDATA[
<p>Does Nostr have a Wikipedia page? I'm wondering how old is it, the people behind it, some stories?</p>
]]></description><pubDate>Thu, 26 Jan 2023 15:54:52 +0000</pubDate><link>https://news.ycombinator.com/item?id=34533320</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=34533320</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=34533320</guid></item><item><title><![CDATA[New comment by milansuk in "The AMD Ryzen 9 7900, Ryzen 7 7700, and Ryzen 5 5 7600 Review: Zen 4 at 65 Watts"]]></title><description><![CDATA[
<p>"we were pretty impressed with the overall performance on offer, despite being limited to a 65 TDP limit. Although power consumption on the CPU package was around 90 W at full load, this is still considerably lower than other processors the trio of 65 W Ryzen 7000 chips was pitted against."<p>This is from the Conclusion page. Still pretty impressive.</p>
]]></description><pubDate>Tue, 10 Jan 2023 12:09:29 +0000</pubDate><link>https://news.ycombinator.com/item?id=34323572</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=34323572</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=34323572</guid></item><item><title><![CDATA[New comment by milansuk in "Ask HN: Why did Frontend development explode in complexity?"]]></title><description><![CDATA[
<p>Sure developers have a hard time handling all that HTML, CSS, Javascript and connections(DOM, etc.) between them, but I see a much bigger problem in browsers. There are actually only a few browsers and most of them are based on chromium(or Webkit). Why? Because web standards are huge!<p>Their complexity is so big, that even great programmers have a hard time creating browsers from scratch. The latest example of this is probably Ladybird[0].
Andreas Kling and others have worked on it for years("Just under 1000 days for a bunch of hackers to build a new JavaScript engine"[1]) and they know it will require way more time to just catch up the Chrome[2].<p>A year ago I started to build SkyAlt[3], which doesn't have anything to do with the web. It's not just a browser, but also an IDE where you can create apps with few lines of code or just drag and drop stuff on canvas.
It's written in C and it's only 25K LOC. Compiling takes a few seconds and binary is under 1MB.
There are tons of features that need to be built, but I like its simplicity(relative to the web).<p>[0] <a href="https://github.com/SerenityOS/serenity/tree/master/Ladybird">https://github.com/SerenityOS/serenity/tree/master/Ladybird</a><p>[1] <a href="https://twitter.com/awesomekling/status/1596783757125898243" rel="nofollow">https://twitter.com/awesomekling/status/1596783757125898243</a><p>[2] <a href="https://twitter.com/awesomekling/status/1595387284035145729" rel="nofollow">https://twitter.com/awesomekling/status/1595387284035145729</a><p>[3] <a href="https://github.com/MilanSuk/SkyAlt">https://github.com/MilanSuk/SkyAlt</a></p>
]]></description><pubDate>Mon, 02 Jan 2023 16:19:28 +0000</pubDate><link>https://news.ycombinator.com/item?id=34219635</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=34219635</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=34219635</guid></item><item><title><![CDATA[New comment by milansuk in "Elizabeth Holmes is sentenced to more than 11 years for fraud"]]></title><description><![CDATA[
<p>I was watching White Collar Advice[0] an hour ago and there was an interesting fact(25m30s) that she didn't work since December 2018. If she drove for Uber, it would probably influence the judge.<p>By the way, the whole video is interesting.<p>[0] <a href="https://www.youtube.com/watch?v=aFFxLwvGLhU" rel="nofollow">https://www.youtube.com/watch?v=aFFxLwvGLhU</a></p>
]]></description><pubDate>Fri, 18 Nov 2022 23:50:28 +0000</pubDate><link>https://news.ycombinator.com/item?id=33664688</link><dc:creator>milansuk</dc:creator><comments>https://news.ycombinator.com/item?id=33664688</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=33664688</guid></item></channel></rss>