Let's Write A Harness!

The same model sits behind Cursor, Claude Code, and most other agents - and yet, each of them produces different results when tasked with the same prompts.

Why? Because what makes them different isn't the brain - it's the harness: the code that gives the model memory, tools, and a way to act on the world.

But despite how much the harness matters, most people still treat it as a black box. In this talk we’re going to change that - by building a fully working harness from scratch to learn how it works and how we can design it to make our agents even more capable.