<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Ackley Will — Insights</title><link>https://ackleywill.com/insights/</link><description>Notes from the workbench: computer vision, deep learning, and AI education.</description><language>en-us</language><item><title>A threshold is a decision</title><description>Explore what changes when the model stays the same, but the rule for acting on its scores changes.</description><link>https://ackleywill.com/insights/experiments/a-threshold-is-a-decision/</link><guid isPermaLink="true">https://ackleywill.com/insights/experiments/a-threshold-is-a-decision/</guid><pubDate>Fri, 16 Oct 2026 12:00:00 GMT</pubDate></item><item><title>Accuracy is a starting point, not a conclusion</title><description>A good overall score can hide the mistakes that matter most. Start by looking at the errors.</description><link>https://ackleywill.com/insights/articles/accuracy-is-a-starting-point/</link><guid isPermaLink="true">https://ackleywill.com/insights/articles/accuracy-is-a-starting-point/</guid><pubDate>Fri, 16 Oct 2026 12:00:00 GMT</pubDate></item><item><title>Fine-tuning starts with a baseline</title><description>Before deciding which layers to unfreeze, give yourself a comparison you can trust.</description><link>https://ackleywill.com/insights/notes/start-with-a-baseline/</link><guid isPermaLink="true">https://ackleywill.com/insights/notes/start-with-a-baseline/</guid><pubDate>Fri, 16 Oct 2026 12:00:00 GMT</pubDate></item></channel></rss>