March 14, 2026 — San Francisco, CA
45-Minute Talk with Live Demo
Jeff has three AI systems. Two didn't work. One does — and now it's building the other two.
21 years at Itron deploying smart meters at scale — 2.3 million devices across San Diego. That infrastructure mindset shaped how Jeff built Sabrina: not as a chatbot or a Second Brain, but as a Chief of Staff that takes action. Morning briefings before he wakes up. Agent swarms that write and test code overnight. A voice interface that turns spoken intent into executed tasks.
The breakthrough was context engineering — the Kai history system that gives every AI session full memory of what came before. Without it, nothing worked. With it, everything connected.
What to Expect
This isn't slides about what could be built. This is a live system, running in production, demonstrated on stage.
Mac menu bar, speak naturally. Watch Sabrina process intent, route to the right system, and create tasks — all from a spoken sentence.
The Chief of Staff pattern running live. Email triage, calendar synthesis, and priority recommendations — generated before the day starts.
Launch 3 parallel agents on stage. Watch them pick up work orders, write code, run tests, and commit — in real time.
Pull up the git log. See overnight commits where Sabrina extended her own capabilities — features she designed, wrote, tested, and deployed herself.
Key Messages
The AI model is the least interesting part. The infrastructure around it — context management, file routing, memory systems — is what makes it work.
The Kai history system was the breakthrough. Structured session memory that survives crashes, persists across tools, and compounds over time.
When the cost of delegating drops below the cost of doing it yourself, everything changes. Sabrina made that inversion real.
Not a prototype. Not a demo. A system that has been running daily — triaging email, managing projects, and building itself — for months.
Two systems failed before one worked. This talk covers the dead ends, the wrong abstractions, and the lessons that only come from shipping and watching things break.
"I built three AI systems. Two didn't work. One does — and now it's building the other two."
— Jeff McCracken, NateCon 2026 submissionAbout the Speaker
Jeff spent 21 years at Itron, culminating as Director of Global Product Management. He led the original $300MM SDG&E smart meter deployment — 2.3 million devices across San Diego — and presented at industry conferences worldwide.
Now an independent consultant on AMI modernization for utilities, Jeff brings the same infrastructure mindset to personal AI systems. Sabrina is what happens when you apply large-scale deployment thinking to your own life.
Jeff publishes at EdgeAMI.ai.