Codex or Claude Code. Cursor or GitHub Copilot. Terminal or IDE. If you're wondering which of these gives you the best outputs in AI-assisted coding, the answer is all of them, or none of them. Maybe you fancy Google's Antigravity instead, or the desktop app versions of any of the above. It hardly matters which you use, because your coding agent is only as good as the context you engineer around it.

In this talk we'll define what context engineering is, how it works, what makes it portable across harnesses and models, and why it's as important as the code itself. The cool kids are all saying context is the new code, and they're right, which is why it deserves the same rigor we've always given the code. We'll cover terms like prompt, harness, loop, and graph engineering, and why your context needs tests, just like your code.

By the end, you'll understand your options for durable context, including custom instructions, custom agents, agent skills, agent hooks, agent plugins, and MCP servers. And you'll see why context engineering, at its core, is software engineering, with a desperate need for humans in the loop. You might not need to read the code anymore, but you really do need to read the context.

This 60 minute talk is best for software engineers, technical PMs, and anyone else using AI coding tools, especially those working with token limits or company-chosen tools and models