Every Day Is Judgment Day

Judgment. Taste. A human in the loop. "It's what we need right now," say the LinkedIn hot takes. So we put a human in the loop to review AI-generated code, AI-written issues, and AI-described pull requests, and we've built a bottleneck: one reviewer, one growing queue, and a senior developer staring into the middle distance, muttering profanities to the dog asleep under the desk.

The problem isn't having a human in the loop. It's that nobody gets a chance to talk about the change until it's merged and someone else's work depends on it.

This talk unpacks what applying judgment to a change means through four questions:

1. Is this the right thing to build?
2. Does it work?
3. Will it hold up?
4. Does it say the right thing?

For each, we'll sort out what automated checks (coding standards, static analysis, tests, accessibility, visual regression) can answer and what needs a person. Some of it needs more than one person: a conversation with a developer peer, a PM, a content editor, or a client.

Because somewhere along the delivery line, you're going to have a conversation about That Change. Whether it's collaborative or confrontational depends on when you have it. Early, over a working per-PR preview, it sounds like "is this what you meant?" Late, it sounds like "why did this ship?"

Whether your pipeline is mature or you've been meaning to build one, you'll leave with a way to layer automated checks and preview environments so most changes are fast to review and easy to approve, and the load spreads across the team instead of piling onto one senior developer. You'll also be able to tell when a change needs a quick look and when it needs a conversation.

Will it solve all your AI-generated problems? No. But it should make life easier for the humans in your loop.