DryLine

DryLine is an AI-native digital twin for production lines. Large factories already run digital twins, but changing one still takes a specialist, so most teams never get to ask "would a fourth tester help?" before they buy it. In DryLine you describe a change in plain words, or start from a photo or floor plan of the line. Meta's models draft the change as a proposal; the server validates it and sets its risk, and a person approves it. A deterministic discrete-event simulation runs the baseline and every what-if, the server ranks them on the same comparison, and the baseline and the what-if play side by side in 2D and 3D on one clock, moved only by each run's sampled trace. On request, the AI tests ideas by itself: Muse Wonder picks each step, DryLine carries it out within fixed limits (3 what-ifs, 4 runs, 20 minutes), and everything the AI makes is labelled; it never publishes. When Muse Spark explains a result, DryLine checks every number it writes against the computed results, and an answer that does not match is not kept. A proof panel shows each run's fingerprints: same inputs, same result. It ships with an electronics assembly-and-test line and four more templates (sneaker factory, bread bakery, corrugated box plant, distribution center), all on synthetic data. Live demo: https://dryline.up.railway.app Judge account: judge@geminus-demo.invalid (Geminus was DryLine's working name) tL.dBo23NnqGNtAspbH-99vj Meta's models propose. DryLine checks. A person decides.