Daily Briefing — Wednesday, November 11, 2026

Data & AI Platforms Briefing

Postgres and ClickHouse are converging into one packaged OLTP-to-OLAP stack, while governance controls (expiring grants, catalog wikis, per-project cost attribution) move into the core platforms that agents read from.

⇣ Jump To

Click any section or topic below to jump to it.

Move & Transform

CDC ·  ELT/ETL Ingestion

Store & Architect

Lakehouses ·  Table Formats ·  Architectural Patterns ·  Query Engines ·  Specialty Platforms

Consume & Activate

AI-Driven Consumption ·  Semantic Layers & Retrieval

Govern & Operate

Orchestration & Workflow ·  Data Observability ·  Catalogs & Metadata ·  Governance, Security & Compliance ·  FinOps for Data

⚡ Quick Takes

Story Signal
↗  ClickHouse Managed Postgres goes GA, and the Xata acquisition brings branching and scale-to-zero OLTP-to-OLAP CDC is turning into a built-in feature of the platform
↗  Airbyte ships an MCP server with a context layer for agent queries over synced data Ingestion tools want to be the agent’s entry point, not just the pipe
↗  Zerobus Ingest into default storage is now generally available Push ingestion is skipping the bus for lakehouse landing
↗  EDB buying Altinity to integrate PostgreSQL with ClickHouse Postgres + ClickHouse is emerging as a packaged HTAP pattern
↗  Databricks’ Commitment to European Sovereignty Sovereign-cloud placement is now a lakehouse selection criterion in the EU
↗  Parquet FINITE_VECTOR logical type passes vote; format PR merged Embeddings get a first-class type at the file-format layer
↗  Iceberg Rust 0.11.0 released with write-path encryption and partial v3 support Non-JVM engines are moving toward Iceberg v3
↗  The largest expansion of ClickHouse yet ClickHouse is now competing as a full platform, not just an engine
↗  AI-energized trillion-row database startup Ocient raises $100 million Trillion-row, cost-sensitive analytics still supports independent vendors
↗  project44 Turns Supply Chain Data Into AI-Powered Decisions with Snowflake Data providers are shipping agents on the customer’s warehouse
↗  Accelerate GPU training on Unity Catalog tables with UCDataset Training reads straight from governed tables, with masking support still missing
↗  How organizations progress from trusted data to trusted AI and agents Agent rollouts are outpacing data readiness
↗  Trigger Lakeflow Jobs when upstream jobs complete (Beta) Native orchestration now covers cross-workspace job dependencies
↗  Announcing ClickHouse Observability — fully managed observability on ClickHouse Cloud Telemetry is being priced as a warehouse workload
↗  Unity Catalog wikis are in Beta Platform catalogs are absorbing documentation and context
↗  DataHub Cloud v2.4 Standalone catalogs are competing on curated context for agents
↗  Apache Polaris 1.9.0 RC1 goes to vote after license fix The open Iceberg catalog is moving toward row and column enforcement
↗  Temporary role grants to users (Preview) Time-bound access is now native to the warehouse
↗  BigQuery compute cost attribution in editions (Preview) Shared-reservation chargeback gets native attribution

Move & Transform

› CDC

ClickHouse — November 2026

ClickHouse Managed Postgres goes GA, and the Xata acquisition brings branching and scale-to-zero

ClickHouse Managed Postgres, the NVMe-backed service with native CDC into ClickHouse through WalShadow and ClickPipes, is now generally available on AWS, with GCP in private preview. ClickHouse is also acquiring Xata for instant branching, cloning and scale-to-zero on copy-on-write storage, and plans a dedicated tier for agent workloads. For platform teams, the operational database and the analytics engine now ship as one vendor with replication built in, so there is no separate CDC tool to run. Note that backup storage and egress become billable at GA.

✍️ Sai Srirampur, ClickHouse · Read article →

› ELT/ETL Ingestion

Blocks & Files — November 2026

Airbyte ships an MCP server with a context layer for agent queries over synced data

Airbyte has released an MCP server that lets users query data in natural language from Claude or ChatGPT. It can run direct queries on Snowflake and BigQuery destinations fed by 50+ source connectors. Ingestion vendors are trying to move up the stack from moving rows to answering questions about them, which raises the question of whose governance applies when an agent reads a destination through the loader.

✍️ Airbyte (via Blocks & Files storage ticker, Nov 6) · Read article →

Databricks Docs — November 2026

Zerobus Ingest into default storage is now generally available

Lakeflow Connect’s Zerobus Ingest, a push-based write path that skips a message bus, can now write into tables on Databricks default (serverless-managed) storage at GA. This removes the requirement to bring your own cloud storage location for push-ingested event data. Teams that stood up Kafka mainly as a landing buffer for the lakehouse now have one more reason to reconsider it.

✍️ Databricks · Read article →

↑ Top

 

Store & Architect

› Lakehouses

Databricks — November 2026

Databricks’ Commitment to European Sovereignty

Databricks commits to launching natively on the AWS European Sovereign Cloud and to bringing Azure Databricks to Microsoft Sovereign Public Cloud regions. It also plans new Azure Italy and Azure Spain regions and offers direct contracting with a Databricks EU entity. These are commitments, not launches: no delivery dates are given. A companion release the next day announced more than $500M of investment in the Netherlands over three years.

✍️ Rory Patterson, Databricks · Read article →

› Table Formats

Apache Parquet (dev list) — November 2026

Parquet FINITE_VECTOR logical type passes vote; format PR merged

The Parquet community passed the FINITE_VECTOR logical type with 7 binding votes and merged the format change on November 10. The type annotates an existing list structure, so older readers just see a list, and it requires finite elements (no NaN, null or infinity) with no new physical layout. This gives embeddings a standard on-disk type in the file format itself, which matters to anyone keeping vectors in the lakehouse instead of a separate vector store.

✍️ Apache Parquet PMC · Read article →

Apache Iceberg (dev list) — November 2026

Iceberg Rust 0.11.0 released with write-path encryption and partial v3 support

Iceberg Rust 0.11.0 adds client-side encryption on the write path and incremental v3 support: row lineage, nanosecond timestamps, delete vectors and Variant in metadata (Variant reads and writes are not there yet). The Rust implementation underpins DataFusion, Polars and several new engines, so v3 coverage there sets how quickly non-JVM engines can adopt v3 tables.

✍️ Apache Iceberg PMC · Read article →

› Architectural Patterns

Blocks & Files — November 2026

EDB buying Altinity to integrate PostgreSQL with ClickHouse

EnterpriseDB is acquiring Altinity, the ClickHouse services and tooling specialist founded in 2017, which will become EDB’s Real-Time Analytics group. Terms were not disclosed. EDB plans integrated enterprise ClickHouse capabilities this quarter, and its roadmap includes a shared Apache Iceberg catalog and consistent access controls across Postgres and ClickHouse. Combined with ClickHouse’s own Postgres GA the same week, Postgres+ClickHouse is becoming a pairing you can buy from two different vendors.

✍️ Chris Mellor, Blocks & Files · Read article →

› Query Engines

ClickHouse — November 2026

The largest expansion of ClickHouse yet

ClickHouse, which says it now has more than 5,000 customers, launched four products in one day. They are Managed Postgres GA (with the Xata acquisition), ClickHouse Observability in public beta, ClickHouse Security (an agentic SIEM on RunReveal) in private preview, and the open-source Langfuse AI Gateway in public beta. ClickHouse is now pitching itself as a full data platform for AI applications rather than a fast OLAP engine, which puts it in more direct competition with Snowflake and Databricks.

✍️ ClickHouse · Read article →

› Specialty Platforms

Blocks & Files — November 2026

AI-energized trillion-row database startup Ocient raises $100 million

Ocient raised a $100M Series C co-led by BOKA Capital and OCA Ventures, bringing total funding to $260M. The company says it expects to triple revenue this year. It sells its OcientAIQ platform mostly direct, with deployments typically in the $2–3M range. The round shows investors still back hyperscale, on-prem-friendly analytics for trillion-row workloads where cloud warehouse economics break down.

✍️ Chris Mellor, Blocks & Files · Read article →

↑ Top

 

Consume & Activate

› AI-Driven Consumption

GlobeNewswire — November 2026

project44 Turns Supply Chain Data Into AI-Powered Decisions with Snowflake

project44’s supply-chain analyst agent, Mo, uses Snowflake Cortex Analyst to turn plain-language questions into SQL against project44 data shared into customer Snowflake accounts. In project44’s own test, Mo got 10 of 10 verified questions right versus 2 of 10 for a general-purpose LLM, and it has answered over 50,000 questions across 210 customers. These are vendor-reported numbers. The case shows that semantic-model-backed text-to-SQL is shipping inside data-provider products, not only internal BI.

✍️ project44 / Snowflake · Read article →

Databricks — November 2026

Accelerate GPU training on Unity Catalog tables with UCDataset

UCDataset, a native PyTorch data loader now in Beta on AI Runtime, streams Unity Catalog Delta tables directly to each GPU without exporting them. In Databricks’ simulated benchmark, a 300 GB feature table on 8 H100s trained about 10x faster than reading an exported Parquet copy. The Beta cannot yet read tables with row filters or column masks, so governed training sets still need a workaround.

✍️ John Spencer et al., Databricks · Read article →

› Semantic Layers & Retrieval

dbt Labs — November 2026

How organizations progress from trusted data to trusted AI and agents

dbt Labs sets out a three-stage maturity model: standardized and scaled, trusted and contextual, unified and interoperable. It argues that agent reliability depends on governed context more than on the model. It cites Fivetran’s 2026 Agentic AI Readiness Index: 41% of organizations run agentic AI in production, but only 15% say they are fully prepared. This is the merged Fivetran–dbt company’s positioning, but the readiness gap is a useful number to bring to planning discussions.

✍️ Daniel Poppy, dbt Labs · Read article →

↑ Top

 

Govern & Operate

› Orchestration & Workflow

Databricks Docs — November 2026

Trigger Lakeflow Jobs when upstream jobs complete (Beta)

A Lakeflow Job can now start when one or more upstream jobs finish, including jobs in other workspaces on the same Unity Catalog metastore. You configure the dependency on the downstream job. Cross-workspace job dependencies are one of the main reasons teams keep an external orchestrator next to Databricks, so this narrows the case for Airflow in Databricks-only setups.

✍️ Databricks · Read article →

› Data Observability

ClickHouse — November 2026

Announcing ClickHouse Observability — fully managed observability on ClickHouse Cloud

ClickHouse Observability, now in public beta, accepts OpenTelemetry logs, traces and metrics at a managed OTLP endpoint, built on ClickHouse Cloud and ClickStack. List pricing is $0.20/GB ingested for logs and traces and $0.10 per million metric points, halved during the beta. For data platform teams already on ClickHouse, pipeline and agent telemetry can now sit in the same engine as the data it describes.

✍️ Mike Shi, ClickHouse · Read article →

› Catalogs & Metadata

Databricks Docs — November 2026

Unity Catalog wikis are in Beta

Data owners and stewards can now draft and publish wikis on tables in Catalog Explorer, written by hand or with Genie Code, covering what the data means, its intended use and its sources. Anyone who can view the table metadata can comment, suggest edits and react. This is Databricks bringing business-glossary-style documentation into the catalog and making it available as agent context, which overlaps with what standalone catalogs sell.

✍️ Databricks · Read article →

DataHub — November 2026

DataHub Cloud v2.4

DataHub Cloud v2.4 continues the Cloud Context rollout: certification of context documents is in public beta with this release. DataHub has also said Cloud Context is going GA. It reports that analytics agents typically start at 40% accuracy or less on benchmark questions without business context, and that Miro’s Snowflake Cortex agent reached 90% once DataHub context was added. This puts DataHub up against platform-native catalogs like Unity Catalog wikis for the same job.

✍️ Manuela Wei, DataHub · Read article →

Apache Polaris (dev list) — November 2026

Apache Polaris 1.9.0 RC1 goes to vote after license fix

Polaris 1.9.0 RC1 opened for a vote on November 11, after RC0 was rejected over incorrect LICENSE contents in the Spark bundle JARs. Current proposals on the list include row- and column-level security built on the Iceberg REST ReadRestrictions concept and chain-aware authorization for views. Those are the features that would make Polaris a real enforcement point rather than just a metadata endpoint.

✍️ Apache Polaris PPMC · Read article →

› Governance, Security & Compliance

Snowflake Docs — November 2026

Temporary role grants to users (Preview)

GRANT ROLE now takes a FOR INTERVAL or UNTIL clause, so a grant expires automatically (for example, GRANT ROLE oncall TO USER alice FOR INTERVAL '24 HOURS'). RESET EXPIRATION adjusts an existing grant, and SHOW GRANTS and GRANTS_TO_USERS gain an expires_on column. Just-in-time access is now built into the platform, so break-glass and on-call grants no longer need external revoke scripts.

✍️ Snowflake · Read article →

› FinOps for Data

Google Cloud — November 2026

BigQuery compute cost attribution in editions (Preview)

BigQuery now attributes compute charges to the query projects that consume an administration project’s reservation capacity (Preview, November 9). The same release made query-performance troubleshooting and job comparison in the jobs explorer generally available. Centralized reservations have always made chargeback hard, and this gives FinOps teams per-project numbers without building their own INFORMATION_SCHEMA allocation models.

✍️ Google Cloud · Read article →

↑ Top

Compiled by Rainvil Labs — Wednesday, November 11, 2026
Sources verified via live web research on November 11, 2026, covering items published November 2–11, 2026 that were not in prior briefings. Outlets used: ClickHouse, Databricks, Snowflake and Google Cloud documentation, dbt Labs, DataHub, Blocks & Files, GlobeNewswire, Airbyte docs, and the Apache Parquet, Iceberg and Polaris mailing lists. Vendor-reported figures are noted as such. This briefing is for informational purposes only and does not constitute legal, regulatory, or investment advice.