The hard lessons of Harvard's failed geoengineering experiment