Open data, followed all the way down to one person inside it.
Every piece here starts with a public dataset and ends with a story. The numbers can be checked: any figure set in mono opens its own evidence in a column to the right, so you never have to leave the sentence you are reading to find out whether it is true.
02 pieces
02 public datasets
271 occupations
4,810 teenagers
work
01 What Makes Writing Writingai & work One dataset measures the same thing twice — once by breaking each job into cognitive abilities, once by rating the job whole. The two answers point in opposite directions, and the job they disagree about most, out of 271, is writing. 0.000 AI-exposure score across 271 occupations · orange = Writers and authors, ranked 4 0.900 → 02 The Line Someone Drewinteractive Before anyone can say “heavy screen users are N times more likely to be depressed”, two lines have to be drawn — and published claims report neither. Move both yourself: the same 4,810 teenagers will give you any answer from 1.15× to 2.97×. total 0 questionnaire totals across 4,810 teenagers · orange = the line, at 14 51 →
files
— Method Four layers, in the same order every time. → — audit.ipynb All 33 published figures re-derived from the raw source, in pandas, without the pipeline. → — verify.py Re-runs both pipelines and refuses to publish anything that drifted. → — 01 · data.json The committed contract for the AI-exposure piece — 271 occupations. → — 01 · build.py Every calculation behind piece 01, including its own robustness audit. → — 02 · data.json The committed contract for the screen-time piece — all 130 precomputed grid cells. → — 02 · build.py Every calculation behind piece 02, and the loop the sliders read from. → — Contact Email, github, colophon. →