Free Lesson

Build a Definition of Done for Claude Code and Codex

Part of The AI-Powered Professional

60 min
Jul 14, 2026 4:00 PM
Virtual (Zoom)

In this video

What you'll learn

Why does your coding agent decide it's "done"?

You'll understand the mechanics of what makes a coding agent stop, and why native confident "done" is a guess it makes

How often do you reply "did you miss anything?

You'll stop interrogating your agent and learn to build a check that runs the real test so it only ever stops on a pass.

What if "done" had to be earned by your agent?

Leave with a tiny harness to wrap Claude Code and Codex that keeps tool calls going until your done criteria passes

Why this topic matters

You already know defining "done" is product's oldest hard problem, the place rework hides. Now your agent calls half-built work "done" with total confidence, and as more agents run longer you can't read every line. Come build a check that makes "done" something a machine verifies, so you stop babysitting and start supervising. That's the shift at the heart of Hardcore Agentic Engineering.

You'll learn from

Greg Ceccarelli

Greg Ceccarelli

Co-Founder • Ex-CPO Pluralsight • Data at GitHub, Dropbox & Google

Previously at

Pluralsight
GitHub
Google
Dropbox
AlixPartners
See all products from Greg