At IBC 2026, Gracenote made a straight-shooting argument that the next phase of AI in TV should start with not asking one giant language model to do everything. The Nielsen-owned content metadata business is betting instead on a combination of domain-specific data, small language models (SLMs), and task-specific agents, with the Model Context Protocol (MCP) acting as the plumbing between them. One question is whether a more focused architecture can scale across the sprawling and notoriously messy metadata requirements of the TV industry? Gracenote launched its Video MCP Server in September 2025, connecting customers’ preferred AI models to its source-verified entertainment metadata. The company has now added a dedicated Sports MCP Server, currently in beta, while developing an entertainment-focused SLM for high-volume workloads where latency and cost become more important. That is a notable shift in tone. Rather than assuming that…