July 2026 Summaries

Cube has introduced Python analysis capabilities, which enhance its existing semantic layer by enabling advanced modeling tasks that go beyond SQL's aggregation capabilities. Traditionally, complex analyses such as forecasting, regression, and clustering required exporting data to a notebook for processing, but with Python analysis, users can attach scripts directly to workbook tabs in Cube. This approach ensures that analyses are performed on consistent, governed data and can be saved, re-run, and shared easily. Users can initiate Python-based analyses through Analytics Chat by requesting specific tasks like forecasting, which are executed when they require statistical or machine-learning libraries. While SQL remains the tool for standard aggregations and time series, Python is employed for more complex analyses, with results rendered directly in the platform. Users can manage scripts within Cube, ensuring transparency and ease of modification, with popular Python libraries preinstalled, and the option to request demos for new users.

The newly released Cube CLI is a single native binary designed to integrate Cube's public REST API directly into the terminal, allowing users to deploy data models and build applications, such as a React dashboard, efficiently. Built on Rust, the CLI offers a complete client interface for Cube's API without using private endpoints, ensuring transparency and reliability. It supports installation across different platforms and uses OAuth 2.0 for secure authentication. The CLI caters to both human users and AI agents by providing features like --json outputs and deterministic exit codes, allowing for seamless automation. The release aims to extend the capabilities of Cube's UI to a public API, promoting a consistent, text-based interface for automation tools and CI pipelines. The demonstration of the CLI's capabilities showed an agent autonomously creating and deploying a data model and React app, highlighting the ease of use and potential for automation. The plan is to continue expanding the CLI and API functionalities, ensuring they remain in sync and serve as a stable foundation for future automation needs.

Cube has introduced a new integration with dbt that allows existing dbt projects to connect with Cube, transforming each dbt model into a governed cube within the semantic layer, which includes dimensions, measures, descriptions, and joins derived from previous modeling efforts. This integration operates by reading dbt models and converting them into foundational cubes in the Cube data model, enabling the creation and extension of these cubes with advanced metrics. Users can manually trigger a pull or automate it via CI pipelines or repository pushes, with Cube regenerating the corresponding cubes and submitting them for review before they go live. The integration maintains dbt as the source of truth for table transformations, while Cube manages the models built on top, facilitating dashboards, embedded analytics, and AI queries without altering the dbt repository. The integration supports a one-way metrics sync from dbt to Cube and is designed to work with various data warehouses like Snowflake, Amazon Redshift, PostgreSQL, and Google BigQuery. The setup involves connecting Cube to the repository via HTTPS or SSH with encrypted credentials, and the process ensures that production data remains untouched by operating in a sandbox environment.

Over the past three decades, the analytics industry has consistently grappled with the challenge of query latency, where the speed at which results are needed often surpasses the efficiency of existing systems. This issue has only intensified with the rise of AI agents, which demand rapid query responses and generate a high volume of queries, thus straining traditional data processing systems. Historically, solutions such as OLAP systems addressed latency by pre-computing aggregates, but these methods faced scalability issues with growing data complexity. Despite advancements in cloud data warehouses, which promised to eliminate the need for pre-computation, pre-aggregations have persisted in various forms, such as materialized views and query result caches. The semantic layer plays a crucial role in optimizing these pre-aggregations by providing the necessary workload insights, enabling efficient data management and query execution. Cube Store has been developed as a high-performance OLAP engine designed to manage pre-aggregations with sub-second latency, leveraging technologies like Apache Arrow and DataFusion for optimal query processing. This system serves as a bridge between data warehouses and users, balancing cost and performance by storing pre-aggregated data in a manner that supports interactive analytics. Recent upgrades to Cube Store have enhanced its capabilities, including support for precise decimals and common table expression planning, which improve its efficiency in handling complex queries and high query volumes typical of AI-driven processes.

The concept of the context layer in data analytics is gaining momentum, distinguished by its role as a comprehensive framework that enables data agents to accurately answer business questions. Unlike the semantic layer, which is integral but specific to executing data through governed metrics and access policies, the context layer encompasses a broader array of descriptive elements such as catalogs, glossaries, and ontologies, which inform but do not execute. The context layer's emergence addresses failures in earlier models where agents struggled with ambiguity in data interpretation, leading to the realization that a missing context layer was essential for effective data interaction. As the industry embraces this layered architecture, the semantic layer remains crucial as the executable core, transforming inquiries into precise and permitted outcomes. In practice, this innovative architecture allows agents to access various tools and resources, assemble necessary context at runtime, and ultimately execute queries through the semantic layer, which ensures that the answers produced are both accurate and compliant. As highlighted by the example of Brex, the integration of a robust semantic layer is indispensable in making AI applications useful by providing a reliable foundation for agents to interpret and act on data insights efficiently.

Cube has introduced Scheduled Tasks, a feature that automates workflows by allowing users to create prompts with specific schedules, eliminating the need for manual initiation. Users can set up tasks with natural-language prompts, choose a schedule—such as hourly, daily, or monthly—and the task will run automatically, executing queries, updating reports, and emailing results. Each task execution generates a chat thread, enabling easy follow-up queries. Tasks can be created through a dialog or by interacting with the Analytics Chat agent, which drafts and saves the instructions. Scheduled tasks run under the security context of their creator, ensuring data access permissions are maintained, and while they cannot edit the data model in this initial release, they can execute predefined workflows like weekly reports. This feature is now available to Cube customers, who can either set tasks themselves or request a demo to explore its full capabilities.

Cube Core version 1.7 introduces significant advancements, highlighted by the general availability of Tesseract, Cube's next-generation data modeling engine, which brings enhanced capabilities like multi-stage calculations and multi-fact views to production. This release also includes comprehensive data modeling improvements, a major performance overhaul of the query orchestrator and SQL API, and substantial changes such as the default activation of Tesseract and the native query pipeline, with several deprecated options removed. Key features involve advanced data modeling tools like multi-stage grain directives, conditional data masking, and improved number formatting, alongside enhanced pre-aggregation options and new API functionalities like client-side Format API and SQL window functions. Additionally, performance optimizations are achieved through a new columnar data transport between JavaScript and Rust, resulting in faster data handling and query processing. The update encourages users to familiarize themselves with the breaking changes and migration paths before upgrading due to several significant defaults and removals in this version.