<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Ted Smoot: the record</title><description>New decisions and falsified tests from tedsmoot.com, as they are logged.</description><link>https://tedsmoot.com/</link><item><title>Decision: Leave a working server alone.</title><link>https://tedsmoot.com/decisions/2026-10-12-leave-a-working-server-alone/</link><guid isPermaLink="true">https://tedsmoot.com/decisions/2026-10-12-leave-a-working-server-alone/</guid><description>I could rebuild my always-on server on the operating system my day job runs on: more practice, one distro everywhere. Chose: Stay put. No rebuild. Why: Nothing was broken. A rebuild would deliver the same machine with its confidence reset to zero, and the practice already happens on a lab box built to be broken.</description><pubDate>Mon, 12 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Falsified: Qwen 3.8 27B can serve as a trusted local auditor.</title><link>https://tedsmoot.com/record#2026-10-05-qwen-3-8-27b-auditor/</link><guid isPermaLink="true">https://tedsmoot.com/record#2026-10-05-qwen-3-8-27b-auditor/</guid><description>It ran 39 searches and still named none of the files it was asked to find. Scored by a script, no model in the loop. Lesson: Activity is not evidence. A claim counts only if a tool call backs it.</description><pubDate>Mon, 05 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Falsified: Top-ranked momentum picks outperform.</title><link>https://tedsmoot.com/record#2026-09-17-momentum-picks/</link><guid isPermaLink="true">https://tedsmoot.com/record#2026-09-17-momentum-picks/</guid><description>41 days, 2,192 assets in a speculative market. The higher the rank, the worse the return. Then the inverted signal failed too. Shut down. Lesson: The score rewarded moves that had already happened. Ranking on them bought the top.</description><pubDate>Thu, 17 Sep 2026 00:00:00 GMT</pubDate></item><item><title>Falsified: Gemma 4 12B can run a multi-step audit on its own.</title><link>https://tedsmoot.com/record#2026-06-07-gemma-4-12b-agent/</link><guid isPermaLink="true">https://tedsmoot.com/record#2026-06-07-gemma-4-12b-agent/</guid><description>It reasoned out the right method, then talked itself in circles and returned an empty answer. Lesson: A capable model is not a capable agent. Evaluate the model and its harness together.</description><pubDate>Sun, 07 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Decision: Take the site down rather than leave the wrong one up.</title><link>https://tedsmoot.com/decisions/2026-05-19-take-the-site-down/</link><guid isPermaLink="true">https://tedsmoot.com/decisions/2026-05-19-take-the-site-down/</guid><description>The old site described a version of my work I had moved past. The newsletter it promoted was going on pause. Chose: A blank page, and a pause on the newsletter. Why: A blank page says less than the wrong page, and nothing on it is false.</description><pubDate>Tue, 19 May 2026 00:00:00 GMT</pubDate></item><item><title>Decision: Write the book that builds, not the one that warns.</title><link>https://tedsmoot.com/decisions/2026-04-13-write-the-book-that-builds/</link><guid isPermaLink="true">https://tedsmoot.com/decisions/2026-04-13-write-the-book-that-builds/</guid><description>I am writing a book on mid-career viability. Two framings surfaced, and the choice shapes everything downstream: the opening chapter, the title, the cover, the marketing, and whether the book is still selling in 2036. I had to commit before the outline could go further. Chose: Build. Roughly 15% diagnosis, 85% prescription. Why: Fear-based books age into embarrassment. Generative ones age into wisdom.</description><pubDate>Mon, 13 Apr 2026 00:00:00 GMT</pubDate></item></channel></rss>