Writing

Notes from the workshop. Architecture decisions, AI experiments, and lessons from building systems that actually run.

2026-03-25

Aegis Falls: Local+Cloud Hybrid AI Architecture

How OpenClaw, a local GPU node, and selective cloud API calls come together into a persistent, memory-enabled AI system. ACT-R cognitive architecture, hybrid inference, and why the split matters.

 2026-03-22

The Aegis Falls Memory Layer

How the Aegis Falls AI system maintains persistent memory using ACT-R cognitive architecture. Activation dynamics, spreading activation, and what it means for an AI to actually remember.

 2026-03-10

ACT-R Cognitive Architecture for AI Agents

Implementing human-inspired reasoning with declarative memory, activation-based retrieval, spreading activation, and procedural productions for autonomous agents.

 2026-03-05

Building Self-Improving Agents

Creating agents that can recursively improve their own prompts and behavior. Six iterations of OpenClaw and what each taught me about agent evolution.

 2026-02-20

Why Local AI Matters

A manifesto on running AI locally. ROCm on AMD hardware, owning your inference stack, and why "if you're not willing to run it locally, you don't own it."

 2026-02-10

Homelab Architecture for AI Development

Proxmox, GPU passthrough, ROCm, and the infrastructure decisions behind building a homelab that actually supports local AI inference and agent development.

 
$ tail -f ~/writing/*.md
---
More posts as I build things.
This is a workshop log, not a blog.