Essays on AI, entrepreneurship, and building in public. One experiment at a time.
My AI agent has the same memory problem as every company I've ever built. Every conversation starts from zero. Every Monday, the organization wakes up with amnesia. Here's the framework that changed how I think about both.
I rewired my entire AI insurance agent to work with any LLM provider. Not because I'm planning to switch. Because the ability to switch is the point.
95% of startups fail. We all know the statistic. What nobody tells you is what it actually feels like to sit with that number at 5 AM, honestly asking yourself which side you're on.
I thought giving my AI insurance agent negotiation superpowers would fix the conversion problem. I was wrong, and the reason I was wrong taught me more about building organizations than the feature itself.
Every competitive advantage built on proprietary software is dissolving. AI doesn't just level the playing field — it removes the field entirely. Here's what replaces it.
What if a single person could run an entire company — not by working harder, but by encoding organizational logic into executable, testable systems? This is the OAC thesis.
I built an AI insurance sales agent. It's in production. And it's painfully, embarrassingly stupid. Here's what that stupidity is teaching me about how organizations actually work.