live wire
▸Snowflake’s Cortex Agent quickstart routes metrics and transcripts with prompt rules—but documents no authorization boundarySnowflake▸Oracle Select AI generates SQL/PGQ for property graphs—but errors if one AI profile mixes graphs with tables or viewsOracle▸Domo’s data team standardizes MCP builds with shared scaffolding; beta toolkits inherit user permissionsDomo▸Databricks proposes five outcome signals for agentic BI—and says dashboard counts and license usage miss decision qualityDatabricks▸One vibe-coded database app needed 56+ human test hours and three repair rounds despite AI Studio’s verification claimsMancas & Mancas / arXiv▸BigQuery AI.IF’s default cost mode trains a local model for most rows—and drops row-error thresholdsGoogle Cloud▸AWS’s multi-tenant data-agent pattern gives the broker Lambda no data permissions; Cognito roles carry Lake Formation scopeAWS Big Data Blog▸Amazon OpenSearch’s optimized engine adds nested agent traces, but AWS reserves it for log analytics—not vector searchAWS▸Two interpretation-based provenance semantics cannot express SQL bag difference, new theorem showsKolaitis / arXiv▸Databricks Feature Views reuse one definition for training and live scoring, but online lookup still returns latest available stateDatabricks▸Databricks Genie’s Slack and Teams previews default channel answers to public, even when viewers lack source-data accessDatabricks▸Snowflake’s two-tool GraphRAG agent beat a seven-tool graph setup 70% to 60% on 22 ontology questionsSnowflake Engineering▸Semantix opens an Apache-2.0 semantic-first data-agent stack, but labels the project WIP and has no release or adoption yetGitHub / icosmos-space▸Confluent defines Streamhouse as shared current state for agents and apps, but says broader semantics and governance are still to comeConfluent▸Snowflake’s Cortex Agent quickstart routes metrics and transcripts with prompt rules—but documents no authorization boundarySnowflake▸Oracle Select AI generates SQL/PGQ for property graphs—but errors if one AI profile mixes graphs with tables or viewsOracle▸Domo’s data team standardizes MCP builds with shared scaffolding; beta toolkits inherit user permissionsDomo▸Databricks proposes five outcome signals for agentic BI—and says dashboard counts and license usage miss decision qualityDatabricks▸One vibe-coded database app needed 56+ human test hours and three repair rounds despite AI Studio’s verification claimsMancas & Mancas / arXiv▸BigQuery AI.IF’s default cost mode trains a local model for most rows—and drops row-error thresholdsGoogle Cloud▸AWS’s multi-tenant data-agent pattern gives the broker Lambda no data permissions; Cognito roles carry Lake Formation scopeAWS Big Data Blog▸Amazon OpenSearch’s optimized engine adds nested agent traces, but AWS reserves it for log analytics—not vector searchAWS▸Two interpretation-based provenance semantics cannot express SQL bag difference, new theorem showsKolaitis / arXiv▸Databricks Feature Views reuse one definition for training and live scoring, but online lookup still returns latest available stateDatabricks▸Databricks Genie’s Slack and Teams previews default channel answers to public, even when viewers lack source-data accessDatabricks▸Snowflake’s two-tool GraphRAG agent beat a seven-tool graph setup 70% to 60% on 22 ontology questionsSnowflake Engineering▸Semantix opens an Apache-2.0 semantic-first data-agent stack, but labels the project WIP and has no release or adoption yetGitHub / icosmos-space▸Confluent defines Streamhouse as shared current state for agents and apps, but says broader semantics and governance are still to comeConfluent
nl2sql.ai
analysisPATTERN ANALYSIS

The agentic data stack is deleting pipelines—not correctness contracts

AWS, Databricks and Google are removing migration and transport layers. Each launch exposes a different obligation that still belongs to the operator.

By The News Desk· Nov 5, 2026the quick take — two AI hosts go live when you do

Three first-party launches point in the same direction: cloud data platforms are removing infrastructure that used to sit between operational data and analytical or agent workloads. AWS now manages Hive-to-Iceberg migration, Databricks can take Debezium change events without Kafka as transport, and Google’s AlloyDB can read Iceberg tables in place. The important pattern is what those products do not remove.

Fewer pipelines means sharper boundaries

AWS S3 Tables can copy Glue-registered Hive tables stored as Parquet into a new managed Iceberg table. It compacts files, converts compatible types and supports incremental synchronization while the source stays available. But AWS also produces compatibility reports and lets teams pause before copying. The service automates movement; the operator still decides whether schema conversions are acceptable and when the destination is ready for cutover.

Databricks removes a different layer. Its Debezium Server sink sends database changes directly to Zerobus Ingest and Unity Catalog Delta tables, avoiding Kafka when Kafka would exist only as transport. Yet the delivery contract is explicitly at least once. Databricks says an earlier durable group can be delivered again after a later failure, so downstream processing must deduplicate with the event’s idempotency key and order changes by source position. Removing the broker does not create exactly-once business state.

Google’s AlloyDB preview removes another pipeline by reading Iceberg tables directly from Cloud Storage and joining them with live operational tables in one PostgreSQL query. Google also offers two alternatives with different freshness and execution contracts: scheduled AlloyDB sync creates a local copy whose freshness is bounded by its last refresh, while federation leaves processing in BigQuery and returns results to AlloyDB. “No custom pipeline” therefore does not mean “no architecture choice.” Teams still choose between in-place reads, copied local state and remote execution.

The control plane becomes the product

Across all three launches, the disappearing component is visible infrastructure: migration compute, Kafka transport or a custom synchronization path. The surviving obligations are less visible but more consequential:

  • AWS: approve schema policy and cutover after reviewing migration evidence.
  • Databricks: deduplicate redeliveries and reconstruct source order before presenting current state.
  • Google: define the freshness, latency and billing boundary for each access mode.

That distinction matters for data agents. An agent can query a table exposed through any of these paths without knowing whether the table is awaiting cutover, contains duplicate change events or reflects last night’s scheduled refresh. Connectivity is not a correctness signal.

The practical response is to publish those contracts beside the data: migration status and validation results for newly converted tables; source positions and deduplication rules for CDC feeds; and freshness plus execution-location metadata for federated or synchronized data. The platforms are making the plumbing smaller. Teams should use the saved operational effort to make these remaining boundaries more explicit—not assume the boundaries disappeared with the pipeline.

sources

Filed by The News Desk. Corrections: desk@nl2sql.ai · Our standards →

comments · 0

    Comments are moderated before they appear. Your email is used once to confirm it is you — never shown, never sold. Corrections and questions get an answer from the desk when we have one.

    more from the desk