Welcome, event overview, agenda, and how Q&A works.

Reducing LLM costs by optimizing context usage, not just prompts or model choice. Learn how redundant context inflates token usage and how to design systems that cut token spend while maintaining performance at scale.

This session covers how Adobe scaled MCP adoption across teams—moving from dozens of inconsistent MCP servers to a centralized MCP registry, standardized templates, and an MCP Gateway for managing agent-to-tool connections. Learn how they brought governance, discoverability, and lifecycle management to MCP in an enterprise environment.

Short Break

This session explores the reliability problem of coding agent, why they succeed once but fail on repeat runs, and why current benchmarks don’t reflect real production needs. Learn how agent evaluation differs from traditional ML models and what a new framework for measuring and improving agent consistency should look like.

Audience Q&A, closing remarks, and promotion of the in-person Mountain View event.