https://samarthya.me/tags/testing/Recent content in Testing on SamarthyaHugo -- gohugo.ioen© Samarthya 2018-2026. All rights reserved.Sat, 07 Nov 2026 18:22:00 -0500https://samarthya.me/posts/testing-ai-code-1-read-before-you-run/Sat, 07 Nov 2026 18:22:00 -0500https://samarthya.me/posts/testing-ai-code-1-read-before-you-run/<div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> You would not plug in a USB stick you found in a car park. Code a model wrote for you deserves the same caution, especially when its job is deleting files. Before running anything, I read the lines that can do damage. </div> <figure><img class="my-0 rounded-md" loading="lazy" decoding="async" fetchpriority="low" alt="The audit as a funnel: 23 generated files and 2,114 lines of Python, 13 search hits for delete or move calls, 11 real calls, all inside two functions, reachable only from the clean and restore commands" src="https://samarthya.me/posts/testing-ai-code-1-read-before-you-run/overview.svg" ><figcaption>From 2,114 lines to two functions worth worrying about.</figcaption></figure> <h2 class="relative group">The app, and why it deserves suspicion <div id="the-app-and-why-it-deserves-suspicion" class="anchor"></div> <span class="absolute top-0 w-6 transition-opacity opacity-0 -start-6 not-prose group-hover:opacity-100 select-none"> <a class="text-primary-300 dark:text-neutral-700 !no-underline" href="#the-app-and-why-it-deserves-suspicion" aria-label="Anchor">#</a> </span> </h2> <p>In <a href="https://samarthya.me/posts/ollama-vs-colibri-beginners-guide/" >an earlier post</a> I ran Qwen3-Coder 30B locally through two engines, Colibri and Ollama. This time I gave it a bigger job: a 200-line specification for <strong>dupfind</strong>, a command-line tool that finds duplicate and near-duplicate files and folders, then helps clean them up. It has three commands:</p>