<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Patrick Rowe — Writing</title><description>Notes on machine-learned potentials, generative models for chemistry, and getting models used.</description><link>https://www.patrickwrowe.com/</link><item><title>What a machine-learned potential actually learns</title><link>https://www.patrickwrowe.com/writing/what-a-potential-learns/</link><guid isPermaLink="true">https://www.patrickwrowe.com/writing/what-a-potential-learns/</guid><description>Descriptors, smoothness, and why transferability is mostly a data problem.</description><pubDate>Tue, 14 Jul 2026 00:00:00 GMT</pubDate></item><item><title>SMILES is a strange language to model</title><link>https://www.patrickwrowe.com/writing/smiles-is-a-strange-language/</link><guid isPermaLink="true">https://www.patrickwrowe.com/writing/smiles-is-a-strange-language/</guid><description>Canonicalisation, invalid strings, and what tokenisation costs you.</description><pubDate>Tue, 09 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Getting a model into a screening pipeline</title><link>https://www.patrickwrowe.com/writing/model-into-a-screening-pipeline/</link><guid isPermaLink="true">https://www.patrickwrowe.com/writing/model-into-a-screening-pipeline/</guid><description>The gap between a good validation number and a decision someone will act on.</description><pubDate>Tue, 19 May 2026 00:00:00 GMT</pubDate></item></channel></rss>