A chart is a claim. This instrument renders one real series two ways, measures the gap between what the chart shows and what the data says, and grades it on Tufte's Lie Factor, the first continuous score for how much a chart lies.
zero-based axis · full window · true orientation
The chart's visual effect matches the data.
Every value is computed live from the source table, exactly as the training labels are. Nothing here is judged by eye.
Twenty-one published charts across eleven mechanisms, including the AI industry's own. Every underlying value is verified against its primary source, and the Lie Factor is computed from the real numbers where a single figure is defined. The scale is signed: below zero an inverted axis reverses a trend, near zero an expanded axis hides one, above three a truncated baseline invents one. One card audits the platform that trained the model.
A corpus no released chart dataset provides: real deceptions, quantified.
Full data and citations in data/real/.