Agentic workflows, LLM orchestration, and production autonomous systems. I architect the tools that think and act — not just the interfaces that pretend to. This page is a record of what I’ve built and what I’m currently thinking about.
Most of what passes for AI strategy is theater. The credible work happens when someone is actually shipping, breaking things in production, and pressure-testing the abstractions against real users.
Five things I built and use. Not concepts, not pilots — running code with real users or a single very demanding one (me).
An OpenClaw-based system that watches the AI industry around the clock — funding rounds, product launches, partnership signals, talent migration. Pulls from a wide source set and synthesizes into briefs.
LIVEAggregates analyst reports, vendor announcements, and market signals across the managed-services landscape. Replaces the “read fifty newsletters a day” tax with a single morning brief.
LIVEOne dashboard for revenue goals, journal entries, concept docs, daily actions, and health metrics. Rebuilt twice — once from a filesystem wipe.
● in useYears of numerology study — Pythagorean and Chinese systems — distilled into a small consumer app. Live on getdaycode.com.
LIVEScrapes job posts, ad spend, and social activity to identify brands actively running UGC campaigns. Surfaces decision-makers — including LinkedIn profiles — for direct outreach.
LIVELong-form pieces written while building, not after the fact. Updated when something is worth saying — never on a content calendar.
built a multi-agent operating system from scratch. Six specialized agents. Cron jobs. A Discord bot named after Opus. Memory layers. Routing logic. The whole stack. I called it
Everyone's obsessed with picking the "best" AI model. The one model to rule them all. Spoiler: it doesn't exist. And the more
You build with something long enough, you stop seeing it as a product and start seeing it as a collaborator. OpenClaw did that for me. Two months in,
A small log I keep open while building. One-liners, ratios, things I want to remember for later.
The hard part of agent design isn’t the agent. It’s the boundary.
If your eval set fits on one screen, you don’t have an eval set.
Most “AI strategy” decks would be improved by a single shipped prototype.
Latency is a feature. So is silence.
Twenty-five years across enterprise tech. Today I build with AI every day — shipping, breaking, and rebuilding the systems most people are still writing decks about.
Notes from building agentic systems. One essay every two or three weeks. No filler.