---
title: "Comment by Kartik N V J K on \"Data Engineering in the Age of GenAI\" | Hashnode"
description: "The line about errors hiding behind clean row counts really lands, because a generated query that runs and returns a plausible number is the most dangerous kind"
url: "https://hashnode.com/posts/data-engineering-in-the-age-of-genai/6a5c751cfbd41eb0e41948bb/comment/6a5f81483de7a18eba619ba5"
---

[Discussion](https://hashnode.com/posts/data-engineering-in-the-age-of-genai/6a5c751cfbd41eb0e41948bb)

Comment [Data Engineering in the Age of GenAI](https://hashnode.com/posts/data-engineering-in-the-age-of-genai/6a5c751cfbd41eb0e41948bb)

[Kartik N V J K](https://hashnode.com/@kartik-nvj)

[Jul 21](https://hashnode.com/posts/data-engineering-in-the-age-of-genai/6a5c751cfbd41eb0e41948bb/comment/6a5f81483de7a18eba619ba5)

The line about errors hiding behind clean row counts really lands, because a generated query that runs and returns a plausible number is the most dangerous kind of wrong. That undocumented business meaning is the real bottleneck, since the model has no way to know a column silently changed units two quarters ago. Have you found a lightweight way to feed that tribal context in, like a data contract or a column-level glossary the model can read?