What problem does it solve? It turns a plain-English description into a reproducible, seeded synthetic-data experiment page, eliminating the manual work of wiring PRNGs, chart renderers, and seed controls for dataset exploration. ## Core Features & Use Cases - Seeded synthetic data generation: A hash-derived PRNG with named sub-streams produces reproducible datasets controllable via a ?seed= query parameter. - Interactive chart rendering: Canvas-based scatter, trendline, grouped bar, run chart, and distribution views with labelled axes, filters, and point hover readouts. - Verification workflow: A model-check script validates domain invariants across a seed census and writes verdict JSON tracking build and verification phases. - Use Case: Ask for an experiment log comparing two treatment groups; receive a single bench/index.html with adjustable effect size, noise, sample size, regression, difference-of-means interval, and CSV export. ## Quick Start Ask the agent to build a seeded data bench page showing a scatter plot with a trendline and group filters from your experiment description.
