<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>AI on Bulrush Labs — Richard Tibbetts</title><link>https://bulrushlabs.com/articles/category/ai/</link><description>Recent content in AI on Bulrush Labs — Richard Tibbetts</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Fri, 02 Oct 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://bulrushlabs.com/articles/category/ai/index.xml" rel="self" type="application/rss+xml"/><item><title>Give the Baseline a Fighting Chance</title><link>https://bulrushlabs.com/articles/2026/10/02/give-the-baseline-a-fighting-chance/</link><pubDate>Fri, 02 Oct 2026 00:00:00 +0000</pubDate><guid>https://bulrushlabs.com/articles/2026/10/02/give-the-baseline-a-fighting-chance/</guid><description>When your research system beats an LLM baseline, check that the baseline got real effort. How to build a strong alternative: make one example work, investigate surprises, spend enough to see stability, and find the work that is still yours.</description></item><item><title>My Coding Agents Have a Dialect</title><link>https://bulrushlabs.com/articles/2026/10/02/my-coding-agents-have-a-dialect/</link><pubDate>Fri, 02 Oct 2026 00:00:00 +0000</pubDate><guid>https://bulrushlabs.com/articles/2026/10/02/my-coding-agents-have-a-dialect/</guid><description>Coding agents have favorite words. I pointed a lab of AI research agents at them: the terms are consistent, family-specific, carry real information, and no human uses them. Plus what that means for pipelines where one model checks another.</description></item></channel></rss>