Developer builds film recommendation site using AI for data tagging, but runs the final application entirely in the browser with no AI at all

The creator spent almost two months using Claude Code to tag a dataset of 10,000 films with custom plot-element tags across about 700 tags total, writing all tags to data files.

The resulting application is a 1.9MB data file of HTML and JSON that runs entirely in users' browsers using simple JavaScript math to compute tag intersections—no backend, database, or server required.

When a user inputs a film like Overboard, the system tallies tag intersections across 10,000 records and surfaces similar films in less than a tenth of a second, producing results comparable to what an LLM would suggest but with minimal computational overhead.