Start Here¶
This site is one long argument, told across ~175 posts and four deep guides: the way you get value from AI is by building infrastructure and method around it — not by prompting harder. Encoded knowledge compounds; improvisation doesn't. And an agent you can't defend is a liability, not an asset.
That thesis has two faces. Skill-Driven Development is how a person or team compounds capability with AI. Defendable Agents is how an agent's decisions become reproducible and evidenced. The Knowledge Context Protocol is the substrate underneath both, and the agentic web is the world they're built for. Compound and defendable are the same infrastructure seen from two angles.
Pick the lane that fits what you're trying to do.
-
I build software with AI
Start with the method, then the receipts.
-
I design systems & architecture
Forty years of architecture, re-examined for the agentic era.
-
I own security, risk, or compliance
Making AI agents defendable to an auditor, a regulator, or a board.
-
I'm just curious where this goes
The big-picture arc, in the author's voice.
Prefer to wander?¶
- The four guides: Knowledge Context Protocol · Skill-Driven Development · Defendable Agents · The Agentic Web
- The writing: the latest posts, or browse by topic
- Lost in the jargon? the Glossary has one-line definitions of everything
- Who's behind this: About Thor — four decades of shipping, Java Champion, co-founder of JavaZone
Everything here is developed in the open. If a post changed your mind — or didn't — I'd like to hear it.