Daily Briefing — Monday, November 9, 2026
A light day: streaming and lakehouse vendors are all converging on one idea — keep live business context queryable for AI agents — while Databricks shows you may not need Kafka to get CDC into the lakehouse.
⇣ Jump To
Click any section or topic below to jump to it.
⚡ Quick Takes
| Story | Signal |
|---|---|
| ↗ Confluent launches Streamhouse, adding a Query pillar | Streaming vendors are moving up the stack into serving; vendor-neutral “Streamhouse” framing vs. lakehouse is the fight to watch. |
| ↗ Databricks: Debezium to Unity Catalog, no Kafka hop | Broker-less CDC into the lakehouse challenges Kafka-as-default transport; idempotency-key design is the part to copy. |
| ↗ Oracle: open lakehouse with semantic layer and Iceberg | Catalog federation is becoming table stakes; Oracle is competing on semantics plus in-place query, not storage. |
Confluent — November 2026
At Current 2026, Confluent (now an IBM company) introduced Streamhouse, an architecture that keeps business context continuously current for AI agents, apps and analytics. Query joins Stream, Connect, Process and Govern as a fifth platform pillar, and Lightning Tables are GA on AWS, with Confluent's internal testing citing roughly 100 ms from creation to serving and 10–100 ms query latency. Worth reading alongside the Streamhouse Working Group (Aiven, Redpanda, StreamNative, Ververica) that positions the term as an open category, not a Confluent trademark.
✍️ Confluent · Read article →
Databricks Engineering — November 2026
Debezium Server 3.7 ships a Zerobus Ingest sink that streams change events straight into Unity Catalog Delta tables with no Kafka broker in the path. NVIDIA-contributed “envelope” mode lands each event in a fixed nine-column table with a SHA-256 idempotency key and source position (e.g. PostgreSQL LSN); delivery is at-least-once, so AUTO CDC dedupes and builds a Type-1 current-state table ordered by LSN. For teams running Kafka purely as a CDC transport, this is a concrete path to drop a hop.
✍️ Victoria Bukta, Omkar Mehta, Wagner Santos · Read article →
SiliconANGLE — November 2026
In an Oracle AI World interview, Oracle's Nipun Agarwal pitches the Autonomous AI Lakehouse as an open layer over existing data rather than a replacement: its catalog can mount others, including Databricks Unity Catalog, Postgres and MySQL, and query in place. Oracle's October 26 announcements added a business semantic layer and broader Apache Iceberg support, and Lake Cache moves hot Iceberg tables into Exadata flash. Note the theCUBE paid-partnership disclosure, and that Oracle has said the semantic layer lands within 12 months.
✍️ Ryan Stevens, SiliconANGLE · Read article →