<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: nylonstrung</title><link>https://news.ycombinator.com/user?id=nylonstrung</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Fri, 15 May 2026 08:44:16 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=nylonstrung" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by nylonstrung in "Launch HN: IonRouter (YC W26) – High-throughput, low-cost inference"]]></title><description><![CDATA[
<p>Unless I misunderstood it seems like this is trailing the pareto frontier in cost and speed.<p>Compare to providers like Fireworks and even with the openrouter 5% charge it's not competitive</p>
]]></description><pubDate>Thu, 12 Mar 2026 19:44:10 +0000</pubDate><link>https://news.ycombinator.com/item?id=47356058</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47356058</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47356058</guid></item><item><title><![CDATA[New comment by nylonstrung in "I stopped using NixOS and went back to Arch Linux"]]></title><description><![CDATA[
<p>It's unfair to act like NixOS just breaks "randomly" or is inherently unreliable<p>It's essentially deterministic and fully reproducible.<p>Issues with Bluetooth, electron etc as described are essentially irrelevant to NixOS and have to do with your configuration</p>
]]></description><pubDate>Wed, 11 Mar 2026 23:54:18 +0000</pubDate><link>https://news.ycombinator.com/item?id=47344267</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47344267</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47344267</guid></item><item><title><![CDATA[New comment by nylonstrung in "I stopped using NixOS and went back to Arch Linux"]]></title><description><![CDATA[
<p>> the benefits are basically minimal/none<p>Sounds like you've never used it. I've daily driven it for ~2 years and would never go back<p>It works great with containers, you can use Nix to build extremely lean OCI images. Mercury uses it this way- the book NixOS in production discusses it</p>
]]></description><pubDate>Wed, 11 Mar 2026 23:41:35 +0000</pubDate><link>https://news.ycombinator.com/item?id=47344119</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47344119</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47344119</guid></item><item><title><![CDATA[New comment by nylonstrung in "Show HN: Swarm – Program a colony of 200 ants using a custom assembly language"]]></title><description><![CDATA[
<p>Appreciate this, very cool of you to do that</p>
]]></description><pubDate>Fri, 06 Mar 2026 10:19:04 +0000</pubDate><link>https://news.ycombinator.com/item?id=47273151</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47273151</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47273151</guid></item><item><title><![CDATA[New comment by nylonstrung in "The Science of Detecting LLM-Generated Text (2024)"]]></title><description><![CDATA[
<p>It sounds like a "cursed problem". Are there any contemporary techniques that show any promise?</p>
]]></description><pubDate>Sun, 01 Mar 2026 08:01:40 +0000</pubDate><link>https://news.ycombinator.com/item?id=47204684</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47204684</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47204684</guid></item><item><title><![CDATA[New comment by nylonstrung in "Tests Are the New Moat"]]></title><description><![CDATA[
<p>Great article. I foresee people rediscovering 'Test Driven Development', probably with a new buzzword slapped on it</p>
]]></description><pubDate>Wed, 25 Feb 2026 22:30:39 +0000</pubDate><link>https://news.ycombinator.com/item?id=47158989</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47158989</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47158989</guid></item><item><title><![CDATA[New comment by nylonstrung in "Mercury 2: Fast reasoning LLM powered by diffusion"]]></title><description><![CDATA[
<p>I changed my mind: this would be perfect for a fast edit model ala Morph Fast Apply
<a href="https://www.morphllm.com/products/fastapply">https://www.morphllm.com/products/fastapply</a><p>It looks like they are offering this in the form of "Mercury Edit"and I'm keen to try it</p>
]]></description><pubDate>Wed, 25 Feb 2026 03:59:49 +0000</pubDate><link>https://news.ycombinator.com/item?id=47147155</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47147155</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47147155</guid></item><item><title><![CDATA[New comment by nylonstrung in "Mercury 2: Fast reasoning LLM powered by diffusion"]]></title><description><![CDATA[
<p>Very good point- if as much energy/money that's gone into ChatGPT style transformer LLMs were put into diffusion there's a good chance it would outperform in every dimension</p>
]]></description><pubDate>Wed, 25 Feb 2026 03:55:29 +0000</pubDate><link>https://news.ycombinator.com/item?id=47147120</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47147120</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47147120</guid></item><item><title><![CDATA[New comment by nylonstrung in "Mercury 2: Fast reasoning LLM powered by diffusion"]]></title><description><![CDATA[
<p>I'm not sold on diffusion models.<p>Other labs like Google have them but they have simply trailed the Pareto frontier for the vast majority of use cases<p>Here's more detail on how price/performance stacks up<p><a href="https://artificialanalysis.ai/models/mercury-2" rel="nofollow">https://artificialanalysis.ai/models/mercury-2</a></p>
]]></description><pubDate>Wed, 25 Feb 2026 01:26:32 +0000</pubDate><link>https://news.ycombinator.com/item?id=47146114</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47146114</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47146114</guid></item><item><title><![CDATA[New comment by nylonstrung in "My lobster lost $450k this weekend"]]></title><description><![CDATA[
<p>It's extremely similar to the fake "agentic" crypto plays a year ago<p>Where Goatseus Maximus and stuff supposedly created coins  and invested autonomously.<p>Obviously it was BS but it fueled a huge amount of attention and speculation</p>
]]></description><pubDate>Tue, 24 Feb 2026 21:38:02 +0000</pubDate><link>https://news.ycombinator.com/item?id=47143487</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47143487</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47143487</guid></item><item><title><![CDATA[New comment by nylonstrung in "Music Discovery"]]></title><description><![CDATA[
<p>I hate these Lovable-generated slopsites</p>
]]></description><pubDate>Mon, 23 Feb 2026 08:03:20 +0000</pubDate><link>https://news.ycombinator.com/item?id=47119391</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47119391</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47119391</guid></item><item><title><![CDATA[New comment by nylonstrung in "0 A.D. Release 28: Boiorix"]]></title><description><![CDATA[
<p>BAR is incredible, probably best RTS right now</p>
]]></description><pubDate>Mon, 23 Feb 2026 08:01:04 +0000</pubDate><link>https://news.ycombinator.com/item?id=47119374</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47119374</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47119374</guid></item><item><title><![CDATA[New comment by nylonstrung in "Agentic Software Engineering Book"]]></title><description><![CDATA[
<p>This is way too dense, you need to distill your thesis and interesting ideas down to a small post if you expect people spending time reading a 417 page PDF</p>
]]></description><pubDate>Mon, 23 Feb 2026 03:51:02 +0000</pubDate><link>https://news.ycombinator.com/item?id=47117862</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47117862</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47117862</guid></item><item><title><![CDATA[New comment by nylonstrung in "A16z partner says that the theory that we’ll vibe code everything is wrong"]]></title><description><![CDATA[
<p>This wasn't a16z monolithically speaking as a firm, it was Anish Acharya talking on a podcast.<p>Seems like he's focused on fintech and not involved in many of their LLM investments</p>
]]></description><pubDate>Sat, 21 Feb 2026 01:47:07 +0000</pubDate><link>https://news.ycombinator.com/item?id=47096617</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47096617</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47096617</guid></item><item><title><![CDATA[New comment by nylonstrung in "Minions – Stripe's Coding Agents Part 2"]]></title><description><![CDATA[
<p>It has all the trappings of NIH syndrome.<p>Reinventing the wheel without explaining why existing tools didn't work<p>Creating buzzwords ("blueprints" "devboxes") for concepts that are not novel and already have common terms<p>Yet they embrace MCP of all things as a transport layer- the one part of the common "agentic" stack that genuinely  sucks and needs to be reinvented</p>
]]></description><pubDate>Fri, 20 Feb 2026 13:23:20 +0000</pubDate><link>https://news.ycombinator.com/item?id=47087736</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47087736</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47087736</guid></item><item><title><![CDATA[New comment by nylonstrung in "The path to ubiquitous AI (17k tokens/sec)"]]></title><description><![CDATA[
<p>Agree that routing is becoming the critical layer here. Vllm iris is really promising for this <a href="https://blog.vllm.ai/2026/01/05/vllm-sr-iris.html" rel="nofollow">https://blog.vllm.ai/2026/01/05/vllm-sr-iris.html</a><p>There's already some good work on router benchmarking which is pretty interesting</p>
]]></description><pubDate>Fri, 20 Feb 2026 13:12:07 +0000</pubDate><link>https://news.ycombinator.com/item?id=47087588</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47087588</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47087588</guid></item><item><title><![CDATA[New comment by nylonstrung in "Closing this as we are no longer pursuing Swift adoption"]]></title><description><![CDATA[
<p>Hard to feel excited for this project when it feels so handwavey and when basic technical decisions have never been nailed down.<p>What are other projects trying something similar that deserve attention?</p>
]]></description><pubDate>Thu, 19 Feb 2026 01:32:36 +0000</pubDate><link>https://news.ycombinator.com/item?id=47068814</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47068814</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47068814</guid></item><item><title><![CDATA[New comment by nylonstrung in "Dario Amodei – "We are near the end of the exponential" [video]"]]></title><description><![CDATA[
<p>Dario is a serial bullshitter and one of the least reliable in the industry when it comes to his predictions.<p>Anthropic is doing good work but he's personally responsible for a good deal of the Irrational Exuberance that plagues the space</p>
]]></description><pubDate>Sun, 15 Feb 2026 23:59:48 +0000</pubDate><link>https://news.ycombinator.com/item?id=47029149</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47029149</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47029149</guid></item><item><title><![CDATA[New comment by nylonstrung in "SCM as a database for the code"]]></title><description><![CDATA[
<p>Trustfall seems really promising for querying files as if they were a db<p><a href="https://github.com/obi1kenobi/trustfall" rel="nofollow">https://github.com/obi1kenobi/trustfall</a></p>
]]></description><pubDate>Sun, 15 Feb 2026 10:11:55 +0000</pubDate><link>https://news.ycombinator.com/item?id=47022555</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=47022555</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47022555</guid></item><item><title><![CDATA[New comment by nylonstrung in "Humanity's Last Programming Language"]]></title><description><![CDATA[
<p>Why make markdown a typed programming language instead of using protobuf/text proto?<p>LLMs already are very familiar with it, there are 100s of protoc plugin to generate code from it, and it's less verbose and more token efficient + testable with protovalidate</p>
]]></description><pubDate>Wed, 11 Feb 2026 03:54:49 +0000</pubDate><link>https://news.ycombinator.com/item?id=46970604</link><dc:creator>nylonstrung</dc:creator><comments>https://news.ycombinator.com/item?id=46970604</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46970604</guid></item></channel></rss>