- prebuild reference files for common genome+annotations and store in s3 - json file will have s3 URIs - json file itself will be s3 URI ... make `--genome` CL argument work with custom json which are themselves located in s3