An "eval," short for evaluation, is a set of structured tests or benchmarks used to systematically assess the performance and quality of an AI model or a program. Creating evals is a foundational practice when building solutions with AI. In this talk Amy Heineike will provide an introduction to evals: What are they? Why do you need them? And how do you get started?

Simon Maple demonstrates how to unlock your AI agent’s full potential. Through live demos, see how structured context, clear specs, and guidance turn agents into reliable teammates that understand your stack and coding style.

A practical tour of Gemini CLI, Google's open-source toolkit for coding agents. Max will unpack its core design principles, from built-in tools and system prompts to memory, context management, and sandboxed environments, and show how developers can tailor and extend it for their own workflows.

Building with MCP servers is both fun and a bit chaotic. The official MCP spec keeps evolving, but agents aren't always quick to catch up, and it isn't always obvious where the gaps are. This talk is part commiseration, and part survival guide. I'll share what we've learned the hard way: from designing tools agents actually use, to experimenting when client support falls short. In a space moving this quickly, hacking and experimenting are our new \"first principles.\"