LacanAin Haze: The Subject Supposed to Code

You've been using Cursor, Claude Code, Copilot — maybe you've even let an agent run unsupervised and come back to something that almost works. But there's a pattern that's quietly eating the agentic coding world and most people haven't tried it yet.

Ralph loops are stupidly simple: a bash while(true) that feeds the same prompt to a coding agent, over and over, with state living in git instead of the context window. Each iteration gets a fresh context. Progress persists in files and commits. The agent picks tasks from a PRD, implements them, and keeps going until it's done.

It sounds like it shouldn't work. It works unreasonably well.

Rick and Mark will explain why, then prove it — live. The audience picks the app. We spec a PRD on the spot. Ralph builds it while we narrate what's happening and why. No pre-baked demos, no safety nets.

You'll walk away understanding how to set up Ralph loops in your own workflow, how to write PRDs that agents can actually execute against, and when to use HITL vs. AFK mode.

Three months ago Rick ranted about Lacan at this same crowd. This is somehow related. Don't worry about it.

Thank you to TekChix for sponsoring this event! We appreciate your support and bringing dinner. And as usual thanks to The Shop for hosting and providing beverages!

Know what’s Happening Next — before everyone else does.

Get App

Host or Publisher LacanAin Haze