The author created a tool called slothsort to manage messy download folders. Slothsort utilizes a local open-weight AI model, Gemma 3 4B via Ollama, to analyze files. It intelligently proposes new folder locations and human-readable names for each file. The tool defaults to a dry run, meaning no files are moved without explicit user confirmation. A one-command undo feature allows users to revert any changes made by slothsort. It is designed to avoid overwriting existing files by appending suffixes to new names. Slothsort also intelligently skips files that are currently being downloaded or have been recently modified. The AI's suggestions are cached, optimizing performance for subsequent runs. The tool degrades gracefully, functioning even without a powerful GPU by utilizing the CPU. It avoids uploading sensitive user data to the cloud, ensuring privacy as files remain on the local machine. Slothsort handles files with no readable text by simply tidying their names, improving efficiency. The entire process of scanning, analyzing, and renaming files can be scheduled to run automatically. This project highlights the value of open-source AI for running sophisticated tasks locally without privacy concerns or recurring costs.