SlopSift

SlopSift is a natural language processing linter designed to detect patterns associated with AI-generated writing. It identifies elements such as canned arguments, unsupported claims, and filler in text by analyzing the grammatical structure rather than relying solely on word lists. The tool uses a custom-trained, compact dependency parser to map tokens, parts of speech, and the relationships between words, enabling it to recognize how sentences make their claims.

Unlike many writing tools that focus on vocabulary or basic parts of speech, SlopSift inspects the deeper structure of sentences. Its approach involves building a dependency graph for each sentence and applying authorable rules to identify structural tells of AI-generated content. The findings are presented in an exact and explainable manner, naming specific issues such as corrective-antithesis, unsupported-certainty, mechanical-outline, vague-attribution, and AI-specific vocabulary.

SlopSift is delivered as a local-first solution, running entirely on the user's device. It can operate in the browser or via command-line interface, ensuring that text analysis stays local and does not require sending data to external servers. The tool is accessible through a live, editable interface where users can input and modify text, with analysis updating in real time. There is also mention of integrating with workflows using commands such as 'bunx slopsift' and analyzing draft files.

SlopSift is positioned within the class of tools focused on AI evaluation and observability, emphasizing local processing and structural analysis over simple word matching.