I need to explore this dataset systematically to find genuinely surprising patterns. Let me start by understanding what would be *expected* versus what's actually there. {"action": "tool_call", "reasoning": "The overview mentions that beer_ratings (85% of data, 14M rows) appears 'isolated' from the product catalog and user registry despite zero join-viable relationships. This is already surprising - a ratings table disconnected from what it rates. Let me verify this claim by checking what foreign keys actually exist in beer_ratings.", "tool": "inspect_columns", "args": {"table": "beer_ratings"}}