Beyond the Red-Green Loop: Building with Coding Agents

​We have spent two years getting better at asking agents for things. The larger opportunity is building the environment they work inside. This talk is about the deterministic scaffolding around a non-deterministic tool: specifications that define both what success looks like and what failure looks like, lint rules invented for the needs of one particular codebase, tests that go well past the unit, accessibility checks that run automatically, review performed by other agents and tightened every time something slips through, and continuous integration as the place where all of it repeats. Each piece exists to convert something you would otherwise have to say into something the system enforces.

About Schalk (our speaker):
/schalkneethling
Schalk Neethling has spent seventeen years building for the open web. Twelve of them were at Mozilla, across MDN Web Docs, Firefox Growth Engineering and developer tooling, including five years leading the MDN frontend through its rewrite to the Yari platform. He now works where that platform depth meets AI-assisted engineering, moving teams from ad hoc prompting to a spec-driven, test-driven practice, where tight feedback loops lift not just the code but the product itself. He is a technical reviewer for O’Reilly and the author of the forthcoming HTML: A Comprehensive Guide. He has worked remotely from Pretoria since 2011.