Coding agents often produce syntactically valid but contextually wrong code for data work because they lack knowledge of your actual schema, governance policies (RBAC, masking, row access), and platform-specific APIs. Generic agents require manual context injection that becomes a maintenance burden. Platform-native agents like Snowflake's CoCo solve this by running under your actual Snowflake role, reading your catalog directly, and shipping with built-in skills for Snowflake-specific tasks like ACCOUNT_USAGE queries, lineage tracing, and PII classification — eliminating the scaffolding overhead.
Table of contents
Coding Agents Belong in Data WorkWhat “Context” Actually Means for a Data StackGovernance Is the Layer Most People Don’t Think About Until It Breaks SomethingGet Luke Hartzell’s stories in your inboxSome Knowledge Can’t Be InjectedThis Is the Problem a Platform-Native Agent Actually Solves1 Impression