Two days with the people building deepest. Accelerate what you already know.
You've shipped with AI. You have a CLAUDE.md. You've felt the leverage.
But you're mostly figuring this out alone. The people around you aren't in it the way you are. The questions you have now aren't answered in blog posts.
Everyone in this room is at your depth. The questions get better when nobody needs the preamble.
The room itself compounds. The conversations between sessions — at lunch, at dinner, in the hallway — are the work, not the break from it. People thinking this far ahead, building this deep, don't find each other often. This is that room.
Hands on keyboards. Real practitioners. No consultants.
Day 1 is about why we're in the room together and what each of us is building. Kevin and Madeline open with what we're seeing and what we hope this summit becomes. Dan Shapiro gives the room a shared language with the 5 Levels — a way to talk about where you actually are without the usual hand-waving. Then everyone goes around: two minutes each, who you are, what you're building, what's got you excited. Day 2 is built on the relationships that start here.
Day 2 is structured as a build stack — each session hands off directly to the next. You'll push your codebase from "agents work sometimes" to "agents operate autonomously," turn your scattered Claude tricks into a system that compounds, and wire it into a workflow your team can run without you in the loop. The conversation with Andrew closes the day because after six hours of building, the best thing to do is zoom out and talk about what it all means.
Level up your CLAUDE.md from "helpful" to "autonomous." Agent rules that actually constrain behavior. A validation harness design. The architecture that lets agents operate unsupervised.
One complete AI-native workflow — spec to ship — on your actual repo. The workflow your team runs on Monday.