<?xml version="1.0" encoding="UTF-8"?>
<rss  xmlns:atom="http://www.w3.org/2005/Atom" 
      xmlns:media="http://search.yahoo.com/mrss/" 
      xmlns:content="http://purl.org/rss/1.0/modules/content/" 
      xmlns:dc="http://purl.org/dc/elements/1.1/" 
      version="2.0">
<channel>
<title>Simon P. Couch</title>
<link>https://simonpcouch.com/blog/</link>
<atom:link href="https://simonpcouch.com/blog/index.xml" rel="self" type="application/rss+xml"/>
<description>A data science blog</description>
<generator>quarto-1.9.36</generator>
<lastBuildDate>Thu, 16 Apr 2026 05:00:00 GMT</lastBuildDate>
<item>
  <title>LLMs running on my laptop can drive coding agents now</title>
  <link>https://simonpcouch.com/blog/2026-04-16-local-agents-2/</link>
  <description><![CDATA[ In December, I wrote a post called “<a href="https://www.simonpcouch.com/blog/2025-12-04-local-agents/">Local models are not there (yet)</a>.” It concluded like so: ]]></description>
  <guid>https://simonpcouch.com/blog/2026-04-16-local-agents-2/</guid>
  <pubDate>Thu, 16 Apr 2026 05:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2026-04-16-local-agents-2/featured.png" medium="image" type="image/png" height="144" width="144"/>
</item>
<item>
  <title>How Posit AI’s Next Edit Suggestions work</title>
  <link>https://simonpcouch.com/blog/2026-03-06-nes/</link>
  <description><![CDATA[ I just loaded some data from the <a href="https://simonpcouch.github.io/forested/">forested package</a> into my R environment. It has a bunch of measurements of forest attributes across Washington State: ]]></description>
  <guid>https://simonpcouch.com/blog/2026-03-06-nes/</guid>
  <pubDate>Fri, 06 Mar 2026 06:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2026-03-06-nes/featured.png" medium="image" type="image/png" height="141" width="144"/>
</item>
<item>
  <title>Introducing Posit AI</title>
  <link>https://simonpcouch.com/blog/2026-03-05-posit-ai/</link>
  <description><![CDATA[ Today we released <a href="https://posit.co/products/ai">Posit AI</a>, an AI service for data scientists. This was a huge effort that spanned many teams over several months, and I’m really excited to have it out in the world. It’s <em>really good.</em> ]]></description>
  <guid>https://simonpcouch.com/blog/2026-03-05-posit-ai/</guid>
  <pubDate>Thu, 05 Mar 2026 06:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2026-03-05-posit-ai/featured.png" medium="image" type="image/png" height="146" width="144"/>
</item>
<item>
  <title>Electricity use of AI coding agents</title>
  <link>https://simonpcouch.com/blog/2026-01-20-cc-impact/</link>
  <description><![CDATA[ Throughout 2025, we got better estimates of electricity and water use of AI chatbots. There are all sorts of posts I could cite on this topic, but a favorite is <a href="https://www.sustainabilitybynumbers.com/p/ai-footprint-august-2025">this blog post</a> from Our World in Data’s Hannah Ritchie. On the electricity front: ]]></description>
  <guid>https://simonpcouch.com/blog/2026-01-20-cc-impact/</guid>
  <pubDate>Tue, 20 Jan 2026 06:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2026-01-20-cc-impact/featured.png" medium="image" type="image/png" height="144" width="144"/>
</item>
<item>
  <title>chores 0.3.0 and local LLMs</title>
  <link>https://simonpcouch.com/blog/2025-12-10-chores-0-3-0/</link>
  <description><![CDATA[ The tl;dr: ]]></description>
  <guid>https://simonpcouch.com/blog/2025-12-10-chores-0-3-0/</guid>
  <pubDate>Wed, 10 Dec 2025 06:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-12-10-chores-0-3-0/featured.png" medium="image" type="image/png" height="144" width="144"/>
</item>
<item>
  <title>Local models are not there (yet)</title>
  <link>https://simonpcouch.com/blog/2025-12-04-local-agents/</link>
  <description><![CDATA[ I understand the appeal of local models. Using coding agents like Claude Code or Codex, it’s not difficult to rack up a hundred dollars of usage in the course of a work week. Besides the price, if you’re working with sensitive IP or confidential data, you need to really believe that providers like Anthropic and OpenAI can be trusted with your data. And then, there’s evil billionaires. What if you could run models that were nearly as good on your own laptop? ]]></description>
  <guid>https://simonpcouch.com/blog/2025-12-04-local-agents/</guid>
  <pubDate>Thu, 04 Dec 2025 06:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-12-04-local-agents/featured.png" medium="image" type="image/png" height="144" width="144"/>
</item>
<item>
  <title>Analyzing my music listening data with Databot</title>
  <link>https://simonpcouch.com/blog/2025-12-03-wrapped-databot/</link>
  <description><![CDATA[ It’s Spotify Wrapped season, which means that everyone I follow on instagram is posting screenshots on their stories and I need to export my iTunes Library metadata as an .xml file and analyze it with the tidyverse. (If you’re new here, I do a little <code>group_by() %&gt;% summarize()</code> on my own music listening data <a href="https://www.simonpcouch.com/blog/2023-11-30-listening-2023/">each</a> <a href="https://www.simonpcouch.com/blog/2022-12-01-listening-2022/">year</a>.) ]]></description>
  <guid>https://simonpcouch.com/blog/2025-12-03-wrapped-databot/</guid>
  <pubDate>Wed, 03 Dec 2025 06:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-12-03-wrapped-databot/featured.png" medium="image" type="image/png" height="144" width="144"/>
</item>
<item>
  <title>When plotting, LLMs see what they expect to see</title>
  <link>https://simonpcouch.com/blog/2025-11-26-bluffbench/</link>
  <description><![CDATA[ <em>This post is a cross-post of a <a href="https://posit.co/blog/introducing-bluffbench/">post</a> on the Posit Blog, co-written with Sara Altman.</em> ]]></description>
  <guid>https://simonpcouch.com/blog/2025-11-26-bluffbench/</guid>
  <pubDate>Wed, 26 Nov 2025 06:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/assets/blank.png" medium="image" type="image/png" height="1" width="1"/>
</item>
<item>
  <title>side::kick(), a coding agent for RStudio</title>
  <link>https://simonpcouch.com/blog/2025-11-11-sidekick/</link>
  <description><![CDATA[ I’m excited to share <a href="https://simonpcouch.github.io/side/"><code>side::kick()</code></a>, an experimental coding agent for RStudio users, built entirely in R. It can interact with your files, talk to your active R session, and run code. ]]></description>
  <guid>https://simonpcouch.com/blog/2025-11-11-sidekick/</guid>
  <pubDate>Tue, 11 Nov 2025 06:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-11-11-sidekick/featured.png" medium="image" type="image/png" height="145" width="144"/>
</item>
<item>
  <title>I’m… writing a newsletter?</title>
  <link>https://simonpcouch.com/blog/2025-10-08-newsletter/</link>
  <description><![CDATA[ Between Positron Assistant, Databot, ellmer, chatlas, and their offshoots, there’s been a lot of LLM-related news coming out of Posit in 2025. Many folks across the organization felt that it was hard to keep up with new developments in the space, both internally and in the wider world. With this in mind, my colleague <a href="https://www.linkedin.com/in/sarakaltman/">Sara Altman</a> and I started working on an internal newsletter in June; released every two weeks, we’d give the most concise possible synopsis of movements within the company as well as the news from the wider space that we were paying the closest attention to. ]]></description>
  <guid>https://simonpcouch.com/blog/2025-10-08-newsletter/</guid>
  <pubDate>Wed, 08 Oct 2025 05:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-10-08-newsletter/featured.png" medium="image" type="image/png" height="141" width="144"/>
</item>
<item>
  <title>I was wrong about tidymodels and LLMs</title>
  <link>https://simonpcouch.com/blog/2025-08-26-predictive/</link>
  <description><![CDATA[ One of my most visceral memories of my first interactions with LLMs was asking that first release of ChatGPT in late 2022 to write code to fit a linear regression with tidymodels. The model hallucinated a function <code>tidymodels::install_tidymodels()</code> again and again. That function does not exist. If it did, there’d be some serious chicken and egg happening. This thing was goofy. ]]></description>
  <guid>https://simonpcouch.com/blog/2025-08-26-predictive/</guid>
  <pubDate>Tue, 26 Aug 2025 05:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-08-26-predictive/featured.png" medium="image" type="image/png" height="143" width="144"/>
</item>
<item>
  <title>R and the Model Context Protocol</title>
  <link>https://simonpcouch.com/blog/2025-08-14-mcptools/</link>
  <description><![CDATA[ This is a <a href="https://www.tidyverse.org/blog/2025/07/mcptools-0-1-0/">cross-post</a> from tidyverse.org. ]]></description>
  <guid>https://simonpcouch.com/blog/2025-08-14-mcptools/</guid>
  <pubDate>Thu, 14 Aug 2025 05:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-08-14-mcptools/featured.png" medium="image" type="image/png" height="143" width="144"/>
</item>
<item>
  <title>How I’m using Claude Code to write R code</title>
  <link>https://simonpcouch.com/blog/2025-07-17-claude-code-2/</link>
  <description><![CDATA[ A couple months ago, I <a href="https://www.simonpcouch.com/blog/2025-03-26-claude-code/">wrote a bit</a> about how I was using Claude Code to help me write R code. At the time, I mostly just shared my impressions of working with the tool and some prompting tips. In the month or two after I wrote the post, my usage waned; I was mostly back to using LLMs only for shorter, more narrowly-scoped tasks. A few weeks ago, though, we put together some tooling that has helped me get much more out of the tool and thus made me interested in using it more often again. ]]></description>
  <guid>https://simonpcouch.com/blog/2025-07-17-claude-code-2/</guid>
  <pubDate>Thu, 17 Jul 2025 05:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-07-17-claude-code-2/featured.png" medium="image" type="image/png" height="146" width="144"/>
</item>
<item>
  <title>Kimi K2 and R Coding</title>
  <link>https://simonpcouch.com/blog/2025-07-14-kimi-k2/</link>
  <description><![CDATA[ It was a hoot and a half of a weekend in the LLM world. A company I hadn’t heard of called Moonshot AI released a model called <a href="https://moonshotai.github.io/Kimi-K2/">Kimi K2</a>. From 30,000 feet: ]]></description>
  <guid>https://simonpcouch.com/blog/2025-07-14-kimi-k2/</guid>
  <pubDate>Mon, 14 Jul 2025 05:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-07-14-kimi-k2/featured.png" medium="image" type="image/png" height="142" width="144"/>
</item>
<item>
  <title>Claude 4 and R Coding</title>
  <link>https://simonpcouch.com/blog/2025-05-27-claude-4/</link>
  <description><![CDATA[ <a href="https://www.anthropic.com/news/claude-4">Claude 4</a> dropped on Thursday! Given that Claude 3.7 Sonnet is my daily driver LLM for R coding, I’ve been excited to poke at it. ]]></description>
  <guid>https://simonpcouch.com/blog/2025-05-27-claude-4/</guid>
  <pubDate>Tue, 27 May 2025 05:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-05-27-claude-4/featured.png" medium="image" type="image/png" height="125" width="144"/>
</item>
<item>
  <title>Evaluating Gemini 2.5 Flash on R coding tasks</title>
  <link>https://simonpcouch.com/blog/2025-05-21-gemini-2-5-flash/</link>
  <description><![CDATA[ Google’s preview of their Gemini 2.5 Pro model has <a href="https://www.simonpcouch.com/blog/2025-05-07-gemini-2-5-pro-new/">really made a splash</a>. The model has become many folks’ daily driver, and I’ve started to see “What about Gemini?” in the comments of each of these blog posts if they don’t explicitly call out the model series in the title. Yesterday, Google announced an update of the preview for Gemini 2.5 Flash, a smaller and cheaper version of 2.5 Pro. ]]></description>
  <guid>https://simonpcouch.com/blog/2025-05-21-gemini-2-5-flash/</guid>
  <pubDate>Wed, 21 May 2025 05:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-05-21-gemini-2-5-flash/featured.png" medium="image" type="image/png" height="144" width="144"/>
</item>
<item>
  <title>Evaluating the new Gemini 2.5 Pro update on R coding</title>
  <link>https://simonpcouch.com/blog/2025-05-07-gemini-2-5-pro-new/</link>
  <description><![CDATA[ The title line of <a href="https://developers.googleblog.com/en/gemini-2-5-pro-io-improved-coding-performance/">Google’s release post</a> on the newest Gemini 2.5 Pro release is “even better coding performance.” Reading this, I was curious whether we’d see a notable increase in performance compared to the last generation on R coding tasks; in <a href="https://www.simonpcouch.com/blog/2025-04-01-gemini-2-5-pro/">an earlier post</a>, I saw that the March release of Gemini 2.5 Pro was a contender with Claude 3.7 Sonnet on <em>An R Eval</em>, a dataset of challenging R coding problems. ]]></description>
  <guid>https://simonpcouch.com/blog/2025-05-07-gemini-2-5-pro-new/</guid>
  <pubDate>Wed, 07 May 2025 05:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-05-07-gemini-2-5-pro-new/featured.png" medium="image" type="image/png" height="140" width="144"/>
</item>
<item>
  <title>Evaluating o3 and o4-mini on R coding performance</title>
  <link>https://simonpcouch.com/blog/2025-04-18-o3-o4-mini/</link>
  <description><![CDATA[ 48 hours after the drop of the GPT 4.1 series of models, a trio of non-reasoning models focused on “real-world developer needs,” OpenAI dropped another set of models, o3 and o4-mini. These two models are the latest generation of thinking models from OpenAI, and they form the backbone of <a href="https://github.com/openai/codex">Codex</a>, a new Claude Code competitor from OpenAI. In short, OpenAI wants market share among developers. ]]></description>
  <guid>https://simonpcouch.com/blog/2025-04-18-o3-o4-mini/</guid>
  <pubDate>Fri, 18 Apr 2025 05:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-04-18-o3-o4-mini/featured.png" medium="image" type="image/png" height="144" width="144"/>
</item>
<item>
  <title>How good are the GPT 4.1 models at writing R code?</title>
  <link>https://simonpcouch.com/blog/2025-04-15-gpt-4-1/</link>
  <description><![CDATA[ Yesterday, OpenAI dropped <a href="https://openai.com/index/gpt-4-1/">a new series of models</a> called GPT 4.1, 4.1 mini, and GPT 4.1 nano. This line from their release post, specifically, caught my eye: ]]></description>
  <guid>https://simonpcouch.com/blog/2025-04-15-gpt-4-1/</guid>
  <pubDate>Tue, 15 Apr 2025 05:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-04-15-gpt-4-1/featured.png" medium="image" type="image/png" height="142" width="144"/>
</item>
<item>
  <title>Introducing chores</title>
  <link>https://simonpcouch.com/blog/2025-04-11-chores/</link>
  <description><![CDATA[ The following is a cross-post of a post I put together for the Posit Blog; you can read that post <a href="https://posit.co/blog/introducing-chores/">here</a>. ]]></description>
  <guid>https://simonpcouch.com/blog/2025-04-11-chores/</guid>
  <pubDate>Fri, 11 Apr 2025 05:00:00 GMT</pubDate>
  <media:content url="https://simonpcouch.com/blog/2025-04-11-chores/featured.png" medium="image" type="image/png" height="144" width="144"/>
</item>
</channel>
</rss>
