<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Llm on Max Halford</title><link>https://maxhalford.github.io/tags/llm/</link><description>Recent content in Llm on Max Halford</description><generator>Hugo</generator><language>en-US</language><managingEditor>maxhalford25@gmail.com (Max Halford)</managingEditor><webMaster>maxhalford25@gmail.com (Max Halford)</webMaster><lastBuildDate>Tue, 21 Jul 2026 21:18:41 +0200</lastBuildDate><atom:link href="https://maxhalford.github.io/tags/llm/index.xml" rel="self" type="application/rss+xml"/><item><title>Will agents replace scripts?</title><link>https://maxhalford.github.io/blog/will-agents-replace-scripts/</link><pubDate>Thu, 02 Jul 2026 00:00:00 +0000</pubDate><author>maxhalford25@gmail.com (Max Halford)</author><guid>https://maxhalford.github.io/blog/will-agents-replace-scripts/</guid><description>&lt;p&gt;I am in favor of coding agents. They&amp;rsquo;re here to stay, so we might as well figure out enjoyable ways to work with them. I&amp;rsquo;m aligned with Charity Majors&amp;rsquo; &lt;a href="https://charitydotwtf.substack.com/p/make-ai-boring-again"&gt;Make AI Boring Again&lt;/a&gt; piece. I&amp;rsquo;m also increasingly convinced they won&amp;rsquo;t replace techies, because the non-techies still need us.&lt;/p&gt;
&lt;p&gt;I stumbled on this Slack thread at work:&lt;/p&gt;
&lt;div align="center" &gt;
&lt;figure style="width: 90%; margin: 0;"&gt;
 &lt;img src="https://maxhalford.github.io/img/blog/will-agents-replace-scripts/slack.png"&gt;
&lt;/figure&gt;
&lt;/div&gt;
&lt;p&gt;At work, we log the hours we spend on each customer, and we report them weekly. It&amp;rsquo;s tedious and error-prone. But it&amp;rsquo;s useful: we want to know where time is going, what to improve, and how to price our services.&lt;/p&gt;</description></item><item><title>Autonomous web scraping with Claude Code</title><link>https://maxhalford.github.io/blog/autonomous-web-scraping-claude-code/</link><pubDate>Mon, 11 May 2026 00:00:00 +0000</pubDate><author>maxhalford25@gmail.com (Max Halford)</author><guid>https://maxhalford.github.io/blog/autonomous-web-scraping-claude-code/</guid><description>&lt;p&gt;Andrej Karpathy&amp;rsquo;s &lt;a href="https://github.com/karpathy/autoresearch"&gt;autoresearch&lt;/a&gt; project made a lot of ripples. I guess it&amp;rsquo;s exciting because a program that modifies itself feels like actual AI. The concept of autonomous programs is not new &amp;ndash; see &lt;a href="https://en.wikipedia.org/wiki/Genetic_programming"&gt;genetic programming&lt;/a&gt; &amp;ndash; but &lt;a href="https://github.com/FeSens/auto-arch-tournament/blob/main/docs/auto-arch-tournament-blog-post.md"&gt;what&lt;/a&gt; &lt;a href="https://shopify.engineering/autoresearch"&gt;we&amp;rsquo;re&lt;/a&gt; &lt;a href="https://github.com/MaxHalford/qthun"&gt;witnessing&lt;/a&gt; &lt;a href="https://github.com/WecoAI/awesome-autoresearch"&gt;now&lt;/a&gt; is somewhat more convincing. We just got a bit closer to Neuromancer&amp;rsquo;s world, with a &lt;a href="https://williamgibson.fandom.com/wiki/Turing_Police"&gt;Turing Police&lt;/a&gt; that&amp;rsquo;s in charge of stopping AIs from rewriting themselves.&lt;/p&gt;
&lt;p&gt;While we wait for AI to become self-conscious, there are mundane problems that can be solved with autoresearch. Web scraping is one of these. Indeed, turning unstructured data into knowledge is an excellent problem for LLMs to solve. But there are nuances to have in mind:&lt;/p&gt;</description></item><item><title>Solving Détrak with brute force</title><link>https://maxhalford.github.io/blog/detrak-solver/</link><pubDate>Mon, 02 Feb 2026 00:00:00 +0000</pubDate><author>maxhalford25@gmail.com (Max Halford)</author><guid>https://maxhalford.github.io/blog/detrak-solver/</guid><description>&lt;p&gt;&lt;a href="https://cdn.1j1ju.com/medias/6a/e0/d8-detrak-regle.pdf"&gt;Détrak&lt;/a&gt; is a simple board game. There&amp;rsquo;s a 5x5 grid, and players need to place 12 domino-like pieces to cover the grid completely. The domino symbols are determined by rolling two dice. The dice are six-sided, and their rolls are shared by all players. Points are scored based on adjacencies of matching symbols.&lt;/p&gt;
&lt;p&gt;
 &lt;img src="https://github.com/MaxHalford/detrak/raw/main/46.jpg" width="60%"&gt;
&lt;/p&gt;
&lt;p&gt;The goal is to find the optimal placement of the pieces to maximize the score based on the rolled symbols. What makes it competitive is that everyone has to work with the same rolls. This involves logical thinking and spatial reasoning. But you also have to juggle luck and risk, because you can&amp;rsquo;t predict the dice rolls.&lt;/p&gt;</description></item><item><title>Do LLMs identify fonts?</title><link>https://maxhalford.github.io/blog/llm-font-identification/</link><pubDate>Wed, 30 Jul 2025 00:00:00 +0000</pubDate><author>maxhalford25@gmail.com (Max Halford)</author><guid>https://maxhalford.github.io/blog/llm-font-identification/</guid><description>&lt;p&gt;&lt;em&gt;Spoiler: &lt;a href="https://maxhalford.github.io/llm-font-recognition/"&gt;not really&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;
&lt;p&gt;&lt;a href="https://www.dafont.com/fr/"&gt;dafont.com&lt;/a&gt; is a wonderful website that contains a large collection of fonts. It&amp;rsquo;s more comprehensive and esoteric than Google Fonts. One of its features is a forum where users can ask for help identifying fonts &amp;ndash; check out &lt;a href="https://www.dafont.com/forum/read/522670/font-identification"&gt;this poor fellow&lt;/a&gt; who&amp;rsquo;s been waiting for over two years and bumped his thread. I thought it would be interesting to see if an LLM could do this task, so I scraped the forum and set up a benchmark.&lt;/p&gt;</description></item></channel></rss>