August 2026 Summaries

Media organizations can no longer rely on publishing high volumes of content for traffic and revenue because generative AI has made plausible content abundant and raised audience expectations, according to Code & Theory leaders Michael Fanuzzi and Rebecca Smith. Their proposed “intelligence layer,” piloted for the upcoming RealClear Crime publication, uses AI to support a newsroom’s discovery, drafting, verification, distribution, and learning processes while keeping editors responsible for editorial decisions. The system is designed to flag unverified claims, show supporting sources, require human review at each stage, and disclose AI assistance alongside named human editors at publication. Rather than treating AI as a replacement for reporting or judgment, the approach positions it as a way to eliminate the blank page and direct reporters toward gaps, uncertainties, and higher-value work. The discussion also argues that durable competitive advantage comes not from faster production but from feedback loops that connect a publisher’s audience behavior, archive, and editorial expertise, while maintaining traceable provenance as facts and sources change over time.

AI is making it easier for people across disciplines to create software and content, but engineering leaders from Code & Theory, Braze, and Sanity argue that rapid generation does not guarantee production-ready, valuable results. They emphasize that judgment, problem definition, verification, and the ability to identify AI failures are becoming more important than raw technical output, while individual contributors increasingly act as managers of AI agents. Although AI changes cost structures through token spending and increased output, return on investment should still be measured by customer and product outcomes rather than feature volume. The speakers also identify structured, governed content in a reliable single source of truth as essential for effective AI automation, since AI depends on accurate and accessible information. They conclude that the largest obstacle is organizational adoption: companies must redesign workflows, shift people from authoring toward reviewing AI-generated recommendations, and overcome resistance to long-standing processes.

Sanity’s August releases introduce six updates focused on expanding content editing, workflow architecture, migration support, AI integrations, and media management. Native tables in the Portable Text Editor now provide structured, schema-validatable, GROQ-queryable cells, reducing the need for third-party table plugins. Functions Invoke allows Sanity Functions to call one another directly, supporting composable workflows, parallel processing, privilege separation, and resource isolation while centralizing shared logic. A standalone visual editing package enables click-to-edit functionality for static and non-React sites, while a new agent skill provides guided, source-specific scaffolding for CMS migrations without fully automating the process. Sanity also added a ChatGPT app integration that supports natural-language document management, GROQ queries, and release operations, and Media Library video versioning now preserves earlier asset versions for rollback. The update also highlights Blackbook’s Claude-powered portfolio agent, which uses Sanity’s structured content and GROQ alongside a JSON-LD knowledge graph to keep its website, SEO data, and conversational interface synchronized.