Now

Last updated: May 2025 · Miami, FL

Building

Deep in the agent runtime layer of LogIQ — the part that decides what signal to look at next during an active incident. The architecture question I keep turning over: how does an agent know when it has enough information to act vs. when it should keep looking? The blast radius of getting that wrong in production is real.

Planning to demo at an upcoming Miami tech event. The feedback loop from showing it to people who’ve actually been paged at 2am is irreplaceable.

In the community

Miami’s tech scene has been energizing in a way I didn’t fully expect. Recently wrapped up the DeepStation Vibe Coding Cohort at The LAB Miami and attended Women in Tech Miami. Both left me with ideas I’m still chewing on.

More of this. The informal conversations at the edges of structured events are often where the most useful thinking happens.

Learning

Working through the production side of AI agents — not the architecture diagrams, but the operational reality: how they fail silently, how you monitor something that’s reasoning rather than executing a fixed path, and what observability even means when the system is making judgment calls.

Most writing on this is either too theoretical or too shallow. The interesting stuff is in the middle.

Writing

Writing here when an idea has sat long enough to be specific about. No fixed cadence. The posts tend to come after I’ve hit something repeatedly and finally have something to say about it that isn’t obvious.

This page is inspired by Derek Sivers’ /now movement — a snapshot of what I’m actually focused on, not a highlight reel.