The author mentions it's more than just parquet which duck "just works" with. They're right. Here's a single query against postgres, parquet, json, csv, and Google Sheets:

    ATTACH 'postgres://your_username_password_host:5432/postgres' AS pgdb (TYPE postgres, READ_ONLY);
    with interests as (
        select *
        from
        read_parquet('https://storage.googleapis.com/duck-demo-data/user_interests.parquet')
    ), user_preferences as (
        select *
        from read_json_auto(
            'https://storage.googleapis.com/duck-demo-data/user_preferences.jsonl',
            format = 'newline_delimited',
            records = true
            )
    ), user_details as (
        SELECT * FROM pgdb.user_details
    ), users as (
        select *
        from read_csv_auto('https://storage.googleapis.com/duck-demo-data/users.csv')
    ), one_more_thing as (
        SELECT *
        FROM read_csv_auto(
            'https://docs.google.com/spreadsheets/export?format=csv&id=1O-sbeSxCpzhzZj5iTRnOplZX-dIAiQJAeIO0mlh2kSU',
            normalize_names=True
            )
    )
    select
    users.user_id,
    users.name,
    interests.interest,
    user_preferences.theme,
    user_preferences.language,
    user_details.hobby,
    one_more_thing.one_more_thing
    from users left join
    interests on users.user_id = interests.user_id left join
    user_preferences on users.user_id = user_preferences.user_id left join
    user_details on users.user_id = user_details.user_id left join
    one_more_thing on users.user_id = one_more_thing.user_id

https://www.definite.app/blog/query-any-ducking-thing