Roundup
Your data engineering agent can show its work
Seven read-only views show what a Domain defines, what an Intent changed, where an asset came from, how the last run went and what state the data is in.
This week your data engineering agent's work became something you can inspect. You can ask what it built, what changed, where it came from, and whether the last run passed.
What got easier
See what your data engineering agent built, without opening the code. Seven read-only views answer what a Domain defines, where an asset comes from, what this Intent changed, which downstream assets it affects, how recent runs went and what state the data is in. They appear as cards in the conversation, beside the work they describe.
Try it: in Studio, open an Intent and ask in the conversation what changed and what is affected.
The productivity claim is that the same team ships more with Vibedata: less time to PR and less time per PR, once the team changes what it focuses on and how it builds, and review compounds as each accepted change shrinks the next review.
More things that make your week easier
Ask the semantic model instead of writing dialect SQL. Your data engineering agent can compile a question against the Domain's canonical semantic model and return the SQL or run it read-only. The result names the sources it rejected, the relations it read and the revision it used. DuckDB and MotherDuck today.
A missing grant now names itself and its approver. Where a Domain runs, the permission row names the grant a user is missing and who can approve it. Connections → Check now works on AWS, and an uncertain refusal says it cannot tell.
Try it yourself
Reconcile two live systems before you cut over. The recipe matches equivalent data across two live systems and keeps every difference visible, so the cutover decision rests on a list of differences you can explain.
Where data engineering is heading
Agentic data engineering delegates a scoped outcome and keeps the gates that matter. The post explains why the agent stops at the pull request, and what that changes about the work you hand over.
An agent team gets faster when each task leaves knowledge behind. Each task should leave something the next one can use. The post explains how a run's context becomes the next run's head start.
New to Vibedata?
Start in the hosted sandbox, or install from your cloud marketplace.
Get the weekly roundup by email
One email a week: what got easier in Vibedata, a recipe to try yourself, and where data engineering is heading.