Rewinder
Analyzing AI accuracy on film knowledge
Exploring how AI models handle questions about movies — from plot details to cast information, awards to production trivia. Ultimately interested in whether LLMs can efficiently fact-check themselves at reasonable cost to apply at scale, and whether we can develop user experience models that pull people into the process of verification rather than away from it.
What This Is
This is not a research project about the accuracy of LLMs or specific LLM models. This is a design project to see if LLM-based annotation can reliably surface issues in LLM responses and provide information to analyze and resolve those issues.
As such, see this as a site primarily to click around in and think about design patterns and solutions, not as a database to answer questions.
If you use this as a database to answer research questions I will come to your work personally and mock you.