A scroll-driven narrative about how AI is changing employment, where the charts advance as you read rather than sitting still beside the text. I wrote 4,076 of the 4,951 lines of JavaScript in the published build, including the layer that maps reading position to the active visualization, both choropleth maps, and the job-vulnerability views.
Wrote scroll-narrative.js, 321 of its 350 lines, which decides which visualization is active at a given scroll position and drives the transition into the next one. It sits on top of the third-party scroll-watcher library rather than replacing it.
Built the United States and world maps end to end, 414 and 353 lines, both entirely mine, sharing one interaction model so state-level and country-level patterns read the same way.
Wrote the combined view, the animated transition between encodings, the scatter, and the scroll variant, roughly 1,390 lines across four files.
Wrote high-ai-opportunity-scroll.js at 874 lines and skill-demand-changes-scroll.js at 456 lines, both entirely mine, plus the light and dark theming.
Five Python notebooks covering preprocessing, state-level analysis, and threat scoring, which determined what the final views would actually show.
Each section advances the visualization instead of swapping it out, so the argument and the data move together rather than competing for attention.
Visualizations hold the left three quarters of the screen while the text column updates alongside, so a chart stays in view while you read about it.
Charts morph rather than cut when the question changes, which keeps the reader oriented instead of forcing them to re-read the axes.
The published write-up covers the sketches, the data decisions, and the revisions made after user testing.
