One round I can remember is a 40-minute vibe-coding session, I built a streaming data ingestion pipeline end to end, covering event consumption, schema validation, error handling, and data normalization for downstream processing.