futures gallery

drop real ingredients into the beaker and watch a 14-metric vector assemble from their verbatim engine reference scores. the blend is a simplified additive preview — the shipped engine estimates per-ingredient concentration first (monte-carlo) — but every score, flag and interaction below is production data, untouched.

ships when — the concentration-aware blender the engine already runs per-product is exposed as an api + inci import. mostly packaging, not research. revenue: formulator seats, per-blend api calls.

what a 12-week outcome view could look like. the routines are built from products the live engine has actually analyzed; the trajectories are illustrative — goodbadgray does not have longitudinal outcome data yet, and says so out loud.

ships when — longitudinal check-ins + a validated outcome measure exist. needs a ~10k-user panel and twelve months of patience. that is exactly why every curve above is labeled illustrative.

ingredient regulation, plotted as time-from-today. everything on this radar is a verified regulation with a primary source; the "collision" panel shows what the live engine already knows about its own 28 analyzed products. click a blip.

ships when — legal-data licensing + per-market monitoring feeds are in place. the ingredient half is live today: ban lists already sit inside every engine dossier. revenue: compliance seats for brands and retailers.

a morning mirror that reads your context and narrates the day's skincare order. the hardware and the sentences are fiction; the facts the sentences cite are production engine data. flip the toggles — the narration recomposes.

ships when — hardware / os partners and ambient context (uv, calendar, routine logs) exist. furthest postcard out. the scoring spine it narrates from is production code today.