710: Simen Svale from Sanity

In this episode of ShopTalk, host Dave Rupert welcomes back Simen Svale, CTO and co-founder of Sanity, to discuss the evolving role of content management systems in the age of AI agents. Svale explains Sanity’s origins as a 'content backend' built out of frustration with rigid, blog-centric CMSs, emphasizing its core value as a structured, JSON-based, real-time content repository that’s agnostic to frontends. The conversation dives deep into how AI agents are reshaping content workflows—transforming Sanity from a headless CMS into a 'doubly headless' system where agents can autonomously manage content across websites, documentation, and even internal tools. Svale highlights the growing challenge of siloed content and agent systems, arguing that centralized, structured content like Sanity’s is essential for preventing agents from hallucinating or acting on fragmented data. He shares personal anecdotes about using agents for everything from code generation to content creation, while cautioning that agents need structured environments and human oversight to avoid dangerous overreach. The episode also explores the rise of MCPs (Model Context Protocols) as a powerful new standard for enabling agents to interact with tools, with Svale praising Sanity’s MCP for its simplicity and real-world utility. He concludes with a vision of the future where agents orchestrate complex workflows across systems, while humans focus on high-level strategy and ethical guardrails. Key takeaways include: 1) Structured content is no longer optional—it’s the foundation for reliable AI agents; 2) MCPs are the new UI layer for agents, enabling seamless tool integration; 3) The future of software isn’t more AI, but smarter orchestration between AI and deterministic systems; 4) Human oversight remains critical—agents need 'devil’s advocate' roles to prevent blind obedience; 5) Tools like Sanity’s full-text search and Ingest’s durable workflows help offload AI from context-heavy tasks. The tone is forward-looking, enthusiastic, and slightly wary—optimistic about AI’s potential but grounded in the reality of its current limitations and risks.