Writing
Notes on building software & AI systems.
Essays on agent architecture, context, and the engineering behind AI-native products.
-
OpenViking and the Context Database Thesis: Why AI Agents Need More Than RAG
AI agents don't need better retrieval. They need better-shaped context — unified, navigable, observable, and evolvable. OpenViking's real contribution is the framing, not the framework.
Read essay
-
OpenClaw vs. Hermes: Two Different Answers to the Agent Architecture Question
OpenClaw compounds control. Hermes compounds behavior. The real comparison isn't which is stronger — it's what each thinks the next bottleneck in agent systems actually is.
Read essay
-
Thin Harness, Fat Skills: What gbrain Taught Me About Building AI Products
The next generation of AI products will not be won by thicker harnesses or smarter prompts. It will be won by thin harnesses and fat skills.
Read essay
-
Beyond Prompt Engineering: Context, Harness, and the Product Architecture of AI Agents
Prompt engineering still matters, but the frontier has moved upward — to context, harness, and the platform architecture around model reasoning.
Read essay
-
The Best and Worst Time to Be an AI Builder: Lessons I Already Learned Before Starting to Build My First Independent Product
Before building my first product, I spent a week trying to find a defensible position in AI-assisted patent work. I didn't find one. But I learned something more valuable about what it actually takes to compete.
Read essay