Real data. Intercity rail times (e.g. London–Paris Eurostar 2h16, Beijing–Shanghai HSR 4h18, Tokyo–Osaka Shinkansen 2h21, New York–Washington Acela 2h45, Madrid–Barcelona AVE 2h30) and flight block times are taken from real published services, not generated at random.

Door-to-door model. Every leg includes the real overhead: flight = city↔airport transfer + 90 min check-in/security + flight + 30 min deplane/bags + arrival transfer; rail = ~30 min in/out of stations + ride; driving = road distance ÷ ~95 km/h. This is exactly why the train often beats the plane on short hops.

Isochrone field. Dijkstra finds the earliest door-to-door time to every city across the multimodal network, then the remaining budget radiates out from each reachable city along roads (urban 28 → highway 78 km/h model), layering into the glowing reach bands you see. Islands (Great Britain, Hong Kong, Taipei, Hokkaido) get no fake roads — they connect only by rail/flight.

Extending. To add a city, drop coordinates and airport-transfer minutes into C at the top of the code, then add real durations to RAIL / FLIGHTS. Five regions are covered already.