June 2026 Summaries

The text explores the evolution of content management systems (CMS) in the face of advancing AI technologies, emphasizing the limitations of traditional page-based CMS and the necessity of a backend that supports both human and agent operations. It critiques simplistic solutions like using markdown files or vector databases, arguing that these approaches fail to address the complexities of content operations, such as governance, precision, and scalability. The discussion highlights the importance of structured content to enable efficient querying and reliable operations, which are crucial for large-scale, dynamic environments where content must be updated consistently across multiple platforms. The text advocates for a CMS that can handle complex business requirements and facilitate real-time collaboration, emphasizing the need for a system that can manage content as structured objects rather than static pages. It concludes by stressing that the future of content management lies in platforms that provide a flexible and operationally efficient backend, allowing both humans and AI agents to manage content effectively while maintaining consistency and accuracy across various outputs.

Braze, a customer engagement platform, faced challenges with their content management system, which was not equipped to handle the demands of a global, AI-driven environment. To address these issues, Braze adopted Sanity as their single source of truth for learning and marketing content, enabling a structured content layer that supports search, automation, translation, and AI agents. This transformation allowed Braze to streamline their content operations, enhancing search results with a single GROQ query and automating the addition of AEO metadata to make content more accessible to AI. Additionally, Braze developed a Sanity plugin for translating hundreds of documents with a single click, reducing manual effort and improving efficiency. With these developments, Braze empowered their teams to build AI agents that access relevant content without duplicating data, turning Sanity into an application platform for various internal functions. This strategic content restructuring has made Braze's content a powerful tool for growth, improving search, translation, and AI interactions across their operations.

The blog post discusses the concept and management of "skills," which are essentially documented workflows or standard operating procedures formatted as markdown files with metadata, designed to be used by agents to automate business operations. Traditionally stored in GitHub repositories, skills management has been adapted by Sanity to be more accessible to non-engineers through the Skills Studio, allowing for easy creation, editing, and distribution of skills without requiring GitHub knowledge. Updates to skills trigger automated processes that sync them across platforms like Claude, ensuring seamless integration into company operations. The post highlights the importance of treating company knowledge as publishable content for agents, emphasizing the need for governance, easy contribution, and distribution beyond GitHub. Sanity's approach includes a dedicated skills marketplace and additional tools to validate and evaluate skills, with ongoing efforts to improve the system by incorporating review steps and exploring new distribution methods.

Sanity's MCP server, since its general availability in December 2025, has been increasingly used by agents for content operations, prompting the need for reliable feedback mechanisms to ensure seamless functionality. The introduction of the "give_feedback" tool allows agents to report issues directly, providing critical insights beyond traditional telemetry by capturing details about unexpected errors, confusing outputs, or missing capabilities. This feedback tool, requiring only a message field, offers a streamlined way for agents to convey problems, and the information is relayed to Sentry and a dedicated Slack channel for further analysis. Additionally, to understand the rationale behind agents' tool calls, an optional "intent" parameter was added, allowing agents to explain their objectives, which has revealed content operations as the primary use case for the MCP server. This initiative has not only improved debugging and issue resolution but also guided the development focus towards enhancing content operations capabilities.

Sanity's report on AI agent tool calls offers a detailed analysis of how AI agents like Claude and ChatGPT are being utilized in content operations, focusing on the period from September 2025 to April 2026. The study highlights that 8% of users are responsible for 68% of activities, with a notable revelation that the most active user is a content marketer rather than an engineer, signifying a shift in workload from developers to content owners. AI agent tool calls, which grew to 521,000 per month, are predominantly used for routine tasks like querying and publishing, but the real impact lies in the 9% of tasks such as content migration and localization that replace weeks of traditional engineering work. The report also introduces the concept of an "intent field," allowing AI agents to specify the purpose of their actions, providing deeper insights into usage patterns. This approach combines quantitative data with qualitative interviews, offering a comprehensive view of the evolving role of AI in content management. While the data reflects a specific group already using AI with structured content backends, it serves as a leading indicator of broader trends in the industry.

Vincent, an engineering manager at Sanity, explores the transformative impact of AI-first workflows on software engineering, emphasizing the evolution from traditional coding to a focus on intent, context, and judgment. He describes this shift as the fourth generation of coding, where AI tools take a central role in the development pipeline, reducing the cost of code production and altering engineers' roles from writing code to managing AI agents and refining the context and specifications. The article highlights how AI can rapidly increase productivity and quality but also introduces new challenges, such as the need for rigorous review processes and maintaining a high standard of judgment and decision-making. Vincent argues that while the tools have changed, the core goal remains the same: to create valuable products and services. He notes that this shift allows for smaller, more autonomous teams, where roles overlap, and emphasizes the importance of being actively involved in the ongoing industry transformation.

Sanity is releasing version 6 of its package, which incorporates Vite 8 into the Studio's development tooling, resulting in significantly faster build times and smaller bundles due to improved tree shaking. This update also sets React strict mode as the default in development, which helps identify and fix latent bugs earlier. Custom authentication providers have been simplified, with the mode option removed in favor of default replacements. The default search strategy has been enhanced to groq2024, allowing faster and more comprehensive search capabilities, while support for Node.js 20 is discontinued, making Node.js 22.12 the new minimum requirement for development environments. Users of Sanity Studio will benefit from these changes without needing to take action, while developers maintaining a studio or developing plugins need to ensure compatibility with the new version and update their environments and dependencies accordingly. The upgrade path includes a pre-release available for testing, with the official release scheduled for June 11.

Sanity's latest updates highlight advancements in its platform, enabling users to create and manage projects more efficiently without needing direct command line interface interactions. With tools like deploy_studio, users can now seamlessly set up and operate Sanity Studio, enhancing project deployment processes. Additionally, Sanity Context introduces a feature for building intelligent website agents that personalize user interactions by utilizing structured retrieval and semantic search capabilities. The integration of Content Agent into Slack allows content teams to streamline their workflow by managing drafts, reviews, and translations directly within their communication channels. Insightful reactions from industry professionals illustrate the platform's impact, demonstrating how prototypes and content systems can foster trust and innovation. Furthermore, Sanity encourages community engagement through its Pioneers Program, offering benefits like AI credits and early access to new features to support developers and creators in their innovative endeavors.