Writing

April 2026

AI Memory Has a Confidence Problem

Three serious memory systems have shipped in four months. The field is converging on an insight: flat memory doesn’t work. It hasn’t converged on what to do about it. Here’s the map.

April 2026

Credential Compression

Credentials were proxies for the translation layer between ability and output. AI tools collapsed that layer. Now the proxy points at nothing — and not everyone was relying on the ability it was supposed to measure.

April 2026

The Full Stack: A Cognitive Operating System Inside Claude Code

Five architectural layers — orchestration skills, empirically validated lenses, epistemic memory with dormancy decay, reflexive observation, and shell hooks — that turn a CLI into a multi-session reasoning system. The complete infrastructure guide.

March 2026

Cognitive Lenses for Claude Code

Eleven cognitive lenses, five functional roles, and an ensemble architecture that wires them into skills — the cognitive plumbing inside my Claude Code setup, and how to build your own.

March 2026

Your AI Keeps a Diary About You. It’s Doing It Badly.

Every AI with persistent memory is building a psychographic profile of you — without confidence tracking, decay, or contradiction tolerance. Epistemic Memory treats every stored belief as a weighted hypothesis, not a settled fact.

March 2026

Claude Code Is Not a Code Editor. Here’s What I Actually Built With It.

Skills, hooks, memory, handoffs, and meta-agents — patterns for turning Claude Code into a general-purpose automation tool.