NDJSON / JSON Lines
"Newline-delimited JSON" — one JSON record per line. Streamable, appendable, partitionable. Not an official standard but a de facto convention.
Used for logs (Logstash, Fluentd, Datadog), event streams, big-data export (BigQuery, S3 Athena), AI training data. Files often end in .jsonl or .ndjson. No comments, no outer array — just {...}\n{...}\n{...}.
Beats CSV on nested data and evolution (add a field without breaking readers). Loses to Parquet/Arrow for analytics — text overhead is 10–20× larger.