Lift 3: Verify & Ship¶

Where We're Starting¶

After Lift 2, you've got a decomposed backlog, skills for consistency, and a manual review process that checks AI output against your acceptance criteria. You've felt the power of criteria-based review — no more vague "looks good" calls. Pass or fail.

But you've also felt the honest tradeoff. Manual review works, but it's slow. Every feature gets walked through by hand. And every time you add something new, there's a nagging question: did that change break something you already verified? You won't know unless you re-check everything — and you don't have time for that.

That's the tension this lift resolves. Your acceptance criteria are already doing double duty as spec and test. Now you'll learn to turn them into automated checks that run themselves — and then ship your verified work to a live URL.

What You'll Learn¶

Why your acceptance criteria in Given/When/Then format are already test specifications — and how to have AI turn them into automated tests
The closed loop: criteria → tests → fail → implement → pass — the cycle that replaces manual re-checking
How visual verification extends your safety net to cover what users actually see
How to deploy your verified application — tests pass, then it ships

Sections¶

Your Criteria Are Already Tests — The acceptance criteria you wrote are test specifications in disguise
The Safety Net — AI-generated tests, the closed loop, and visual verification
Ship It — Tests gate deployment — verify it, then ship it to a live URL

By the End of This Lift¶

You can explain why manual review doesn't scale — and what the two-week cliff looks like
You understand that Given/When/Then acceptance criteria map directly to automated test structure
You've had AI generate tests from acceptance criteria and seen them run
You understand visual verification as a way to check what users see, not just what code does
You've deployed your observation network to a live URL
You're ready for Run 3 with automated tests as your safety net