<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: rosmine</title><link>https://news.ycombinator.com/user?id=rosmine</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Fri, 22 May 2026 23:05:34 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=rosmine" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by rosmine in "Was my $48K GPU server worth it?"]]></title><description><![CDATA[
<p>My first advice is to have a test set with clear improvement, and a clear "wow" demo use case. There are lots of "breakthroughs" that seem good but aren't (e.g. some new architecture that doesn't mask past tokens correctly and leaks information), so people will assume it is wrong. To prevent this, you need to be extremely rigorous in your launch materials. If you can make it into a product that people can try out themselves, that goes a long way. You don't need to open source any code (I haven't yet) if people can try it out some other way like a demo website. Good luck! Ping me if you want to chat more</p>
]]></description><pubDate>Fri, 22 May 2026 01:38:48 +0000</pubDate><link>https://news.ycombinator.com/item?id=48230954</link><dc:creator>rosmine</dc:creator><comments>https://news.ycombinator.com/item?id=48230954</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48230954</guid></item><item><title><![CDATA[New comment by rosmine in "Was my $48K GPU server worth it?"]]></title><description><![CDATA[
<p>I agree :) Also, I heard Teknium trained the original Hermes model on 2x 4090. You can do a lot with a little compute</p>
]]></description><pubDate>Thu, 21 May 2026 20:35:38 +0000</pubDate><link>https://news.ycombinator.com/item?id=48228543</link><dc:creator>rosmine</dc:creator><comments>https://news.ycombinator.com/item?id=48228543</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48228543</guid></item><item><title><![CDATA[New comment by rosmine in "Was my $48K GPU server worth it?"]]></title><description><![CDATA[
<p>Yes! It scared me too. I tried to insure it under my renter's insurance policy, but they not surprisingly refused. I had to get business insurance to cover it</p>
]]></description><pubDate>Thu, 21 May 2026 20:31:24 +0000</pubDate><link>https://news.ycombinator.com/item?id=48228499</link><dc:creator>rosmine</dc:creator><comments>https://news.ycombinator.com/item?id=48228499</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48228499</guid></item><item><title><![CDATA[New comment by rosmine in "Was my $48K GPU server worth it?"]]></title><description><![CDATA[
<p>Hi! Thank you so much for posting this! I got back luck/timing when I tried, so happy it made it to the front page! (I am the author)</p>
]]></description><pubDate>Thu, 21 May 2026 20:28:27 +0000</pubDate><link>https://news.ycombinator.com/item?id=48228465</link><dc:creator>rosmine</dc:creator><comments>https://news.ycombinator.com/item?id=48228465</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48228465</guid></item><item><title><![CDATA[Show HN: An LLM that's better at writing]]></title><description><![CDATA[
<p>Standard LLM training is surprisingly bad at making the model outputs match the training data distribution, so the writing quality is bad.<p>I made a new training algorithm called Distribution Fine Tuning (DFT) to fix this.<p>The demo lets you try out a model trained with DFT.<p>More details in the technical report: <a href="https://rosmine.ai/2026/05/18/fixing-llm-writing-with-distribution-fine-tuning/" rel="nofollow">https://rosmine.ai/2026/05/18/fixing-llm-writing-with-distri...</a></p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=48183396">https://news.ycombinator.com/item?id=48183396</a></p>
<p>Points: 3</p>
<p># Comments: 0</p>
]]></description><pubDate>Mon, 18 May 2026 18:18:12 +0000</pubDate><link>https://dft.rosmine.ai/</link><dc:creator>rosmine</dc:creator><comments>https://news.ycombinator.com/item?id=48183396</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48183396</guid></item><item><title><![CDATA[Was my $48K GPU server worth it?]]></title><description><![CDATA[
<p>Article URL: <a href="https://rosmine.ai/2026/05/13/was-my-48k-gpu-worth-it/">https://rosmine.ai/2026/05/13/was-my-48k-gpu-worth-it/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=48122746">https://news.ycombinator.com/item?id=48122746</a></p>
<p>Points: 6</p>
<p># Comments: 0</p>
]]></description><pubDate>Wed, 13 May 2026 14:53:59 +0000</pubDate><link>https://rosmine.ai/2026/05/13/was-my-48k-gpu-worth-it/</link><dc:creator>rosmine</dc:creator><comments>https://news.ycombinator.com/item?id=48122746</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48122746</guid></item><item><title><![CDATA[Show HN: A new test for if your LLM is subtly manipulating you]]></title><description><![CDATA[
<p>Article URL: <a href="https://rosmine.ai/2025/12/10/a-new-test-for-if-your-llm-is-subtly-manipulating-you/">https://rosmine.ai/2025/12/10/a-new-test-for-if-your-llm-is-subtly-manipulating-you/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=46224622">https://news.ycombinator.com/item?id=46224622</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Wed, 10 Dec 2025 22:11:11 +0000</pubDate><link>https://rosmine.ai/2025/12/10/a-new-test-for-if-your-llm-is-subtly-manipulating-you/</link><dc:creator>rosmine</dc:creator><comments>https://news.ycombinator.com/item?id=46224622</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46224622</guid></item><item><title><![CDATA[Why this AI thinks the xfinity logo is a hate symbol]]></title><description><![CDATA[
<p>Article URL: <a href="https://rosmine.ai/2025/01/30/why-this-ai-thinks-the-xfinity-logo-is-a-hate-symbol/">https://rosmine.ai/2025/01/30/why-this-ai-thinks-the-xfinity-logo-is-a-hate-symbol/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=42880518">https://news.ycombinator.com/item?id=42880518</a></p>
<p>Points: 2</p>
<p># Comments: 2</p>
]]></description><pubDate>Thu, 30 Jan 2025 18:20:30 +0000</pubDate><link>https://rosmine.ai/2025/01/30/why-this-ai-thinks-the-xfinity-logo-is-a-hate-symbol/</link><dc:creator>rosmine</dc:creator><comments>https://news.ycombinator.com/item?id=42880518</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=42880518</guid></item></channel></rss>