Mar 28, 2026
Why Your AI Agent Needs a Security Audit Before It Needs More Features
The first question inside a Fortune 50 healthcare org is never 'what can it do' — it's 'what data can it access and who's watching.'
Writing
Direct, technical, anti-hype. Builder perspective on healthcare AI, interface-first architecture, and shipping culture.
Mar 15, 2026
Interface-First Architecture: Stop Coupling Your UX to Your AI Provider
We switched from OpenAI to Claude mid-sprint and nobody noticed. Here's the abstraction layer that made it possible.
Feb 28, 2026
Paperclip: What I Learned Letting an AI Agent Ship Production Code
Paperclip has shipped 47 PRs to RallyDues. 3 had bugs. Here's what the agent gets right, what it misses, and why the three-layer instruction architecture was the breakthrough.
Feb 10, 2026
ROAI: The Only AI Metric That Matters
Your demo is impressive. Your agent writes poetry. Cool. How many hours did it save this quarter?
Jan 22, 2026
Building Multi-Tenant AI: RLS, Isolation, and the Mistakes I Made
AgentFlow serves multiple organizations from one deployment. Here's how I designed the data isolation layer and the three RLS bugs that almost shipped.