<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: grohan</title><link>https://news.ycombinator.com/user?id=grohan</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Fri, 11 Sep 2026 16:57:53 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=grohan" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by grohan in "Shopify acquires Tailwind"]]></title><description><![CDATA[
<p>"load-bearing" - lol, nice easter egg. no one likes the AI-isms in writing!</p>
]]></description><pubDate>Wed, 09 Sep 2026 23:52:36 +0000</pubDate><link>https://news.ycombinator.com/item?id=49636305</link><dc:creator>grohan</dc:creator><comments>https://news.ycombinator.com/item?id=49636305</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49636305</guid></item><item><title><![CDATA[New comment by grohan in "How well do agents use test/verification techniques?"]]></title><description><![CDATA[
<p>Pretty good point - I would actually argue that the imperative nature of prompting takes some of the blame here too. Your initial goal gets "steered" by the agent system and follow-up messages take precedence.</p>
]]></description><pubDate>Tue, 08 Sep 2026 12:19:54 +0000</pubDate><link>https://news.ycombinator.com/item?id=49609319</link><dc:creator>grohan</dc:creator><comments>https://news.ycombinator.com/item?id=49609319</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49609319</guid></item><item><title><![CDATA[Killing Coding Agent Slop with Adversarial Self-Play]]></title><description><![CDATA[
<p>Article URL: <a href="https://usetelos.ai/blog/killing-coding-agent-slop">https://usetelos.ai/blog/killing-coding-agent-slop</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=48924445">https://news.ycombinator.com/item?id=48924445</a></p>
<p>Points: 3</p>
<p># Comments: 0</p>
]]></description><pubDate>Wed, 15 Jul 2026 17:41:11 +0000</pubDate><link>https://usetelos.ai/blog/killing-coding-agent-slop</link><dc:creator>grohan</dc:creator><comments>https://news.ycombinator.com/item?id=48924445</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48924445</guid></item><item><title><![CDATA[New comment by grohan in "Ctoc: Cloc, but for Claude Token Counts"]]></title><description><![CDATA[
<p>Here's a reverse engineer of the Claude token counter API as well as the model's vocabulary. Fast, offline, 96% accurate.</p>
]]></description><pubDate>Wed, 11 Feb 2026 00:16:38 +0000</pubDate><link>https://news.ycombinator.com/item?id=46969043</link><dc:creator>grohan</dc:creator><comments>https://news.ycombinator.com/item?id=46969043</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46969043</guid></item><item><title><![CDATA[Ctoc: Cloc, but for Claude Token Counts]]></title><description><![CDATA[
<p>Article URL: <a href="https://grohan.co/2026/02/10/ctoc/">https://grohan.co/2026/02/10/ctoc/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=46969042">https://news.ycombinator.com/item?id=46969042</a></p>
<p>Points: 1</p>
<p># Comments: 1</p>
]]></description><pubDate>Wed, 11 Feb 2026 00:16:38 +0000</pubDate><link>https://grohan.co/2026/02/10/ctoc/</link><dc:creator>grohan</dc:creator><comments>https://news.ycombinator.com/item?id=46969042</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46969042</guid></item><item><title><![CDATA[New comment by grohan in "Compressed filesystems à la language models"]]></title><description><![CDATA[
<p>Bellard has trained various models, so it may not be the specific 169M parameter LLM, but his Transformer-based `nncp` is indeed #1 on the "Large Text Compression Benchmark" [1], which correctly accounts for both the total size of compressed enwik9 + decompresser size (zipped).<p>There is no unfair advantage here. This was also achieved in the 2019-2021 period; it feels safe to say that Bellard could have likely pushed the frontier <i>far further</i> with modern compute/techniques.<p>[1] <a href="https://www.mattmahoney.net/dc/text.html" rel="nofollow">https://www.mattmahoney.net/dc/text.html</a></p>
]]></description><pubDate>Thu, 27 Nov 2025 04:19:30 +0000</pubDate><link>https://news.ycombinator.com/item?id=46065597</link><dc:creator>grohan</dc:creator><comments>https://news.ycombinator.com/item?id=46065597</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46065597</guid></item><item><title><![CDATA[Compressed filesystems à la language models]]></title><description><![CDATA[
<p>Article URL: <a href="https://grohan.co/2025/11/25/llmfuse/">https://grohan.co/2025/11/25/llmfuse/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=46058033">https://news.ycombinator.com/item?id=46058033</a></p>
<p>Points: 67</p>
<p># Comments: 14</p>
]]></description><pubDate>Wed, 26 Nov 2025 15:03:09 +0000</pubDate><link>https://grohan.co/2025/11/25/llmfuse/</link><dc:creator>grohan</dc:creator><comments>https://news.ycombinator.com/item?id=46058033</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46058033</guid></item><item><title><![CDATA[New comment by grohan in "Fire-Flyer File System from DeepSeek"]]></title><description><![CDATA[
<p>They appear to have Python bindings which seems reasonable from an API / usability perspective? <a href="https://github.com/deepseek-ai/smallpond">https://github.com/deepseek-ai/smallpond</a><p>In terms of fast FUSE - also my first question, appears to be`io_uring` + FUSE :)<p><a href="https://github.com/deepseek-ai/3FS/blob/main/src/lib/api/UsrbIo.md">https://github.com/deepseek-ai/3FS/blob/main/src/lib/api/Usr...</a></p>
]]></description><pubDate>Fri, 28 Feb 2025 03:42:31 +0000</pubDate><link>https://news.ycombinator.com/item?id=43201470</link><dc:creator>grohan</dc:creator><comments>https://news.ycombinator.com/item?id=43201470</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=43201470</guid></item><item><title><![CDATA[Technical Debt Is Entropy in Software]]></title><description><![CDATA[
<p>Article URL: <a href="https://grohan.co/2024/12/27/entropy/">https://grohan.co/2024/12/27/entropy/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=42534757">https://news.ycombinator.com/item?id=42534757</a></p>
<p>Points: 2</p>
<p># Comments: 0</p>
]]></description><pubDate>Sat, 28 Dec 2024 21:09:21 +0000</pubDate><link>https://grohan.co/2024/12/27/entropy/</link><dc:creator>grohan</dc:creator><comments>https://news.ycombinator.com/item?id=42534757</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=42534757</guid></item><item><title><![CDATA[New comment by grohan in "The Annual Cost of Technical Debt: $1.52T"]]></title><description><![CDATA[
<p>Impressive numbers. Does anyone have any read or anecdotes on how much a small/mid/large company loses from low quality of software/bad practices (or conversely profits from the opposite)?<p>Seems like a challenging metric to measure, but always been curious on what the numbers look like.</p>
]]></description><pubDate>Thu, 25 Jan 2024 22:51:28 +0000</pubDate><link>https://news.ycombinator.com/item?id=39136540</link><dc:creator>grohan</dc:creator><comments>https://news.ycombinator.com/item?id=39136540</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=39136540</guid></item><item><title><![CDATA[Factorials and Fun with Vim]]></title><description><![CDATA[
<p>Article URL: <a href="https://grohan.co/2023/08/10/vim/">https://grohan.co/2023/08/10/vim/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=37112821">https://news.ycombinator.com/item?id=37112821</a></p>
<p>Points: 2</p>
<p># Comments: 0</p>
]]></description><pubDate>Sun, 13 Aug 2023 18:43:13 +0000</pubDate><link>https://grohan.co/2023/08/10/vim/</link><dc:creator>grohan</dc:creator><comments>https://news.ycombinator.com/item?id=37112821</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=37112821</guid></item></channel></rss>