Now

Updated: February 12, 2026 Location: Nashville / Murfreesboro, TN

Agent Memory Infrastructure: Built a local GPU-accelerated memory system for my AI agent using SurrealDB (document + graph + vector) and Qwen3-Embedding-4B running on an RTX 5070 Ti. Vector search, concept graphs, temporal edges — all localhost, all mine. Wrote it up here. Published the sanitized prototype on GitHub.
Nosh: Created an open spec for embedding structured, machine-readable content in your page's <head> so AI agents get typed knowledge instead of parsing HTML. 4.1x fewer tokens, same knowledge. Blog post. Spec on GitHub.
bold.casa Redesign: Rebuilt the blog's visual identity — monospace typography, navy blueprint aesthetic, surgical CSS overrides on the tabi theme. Mobile-tested.
Clawguard: Agent ops guardrail layer — budget enforcement, watchdog, shield, control plane. Primary product wedge. Architecture phase.
Legal AI Research: Exploring domain-specific legal embedding models (voyage-law-2, LEGAL-BERT) for evidence analysis and potential pro se legal AI product.

Agent memory architectures: Surveyed MemGPT/Letta OS, OpenAI's approach, Claude's memory, AI Toolkits patterns. Landed on SurrealDB multi-model as the foundation.
Legal AI: voyage-law-2 tops MTEB legal benchmarks. Open-source alternatives exist. Exploring whether pro se legal AI could be the shippable product.
GEO (Generative Engine Optimization): How to make content AI-agent-friendly. Nosh is my answer.

Machine: ROG Strix G16 — Ryzen 9 8940HX, 64GB DDR5, RTX 5070 Ti 12GB, Fedora Silverblue
Stack: OpenClaw + Kit (Claude Opus 4.6), SurrealDB, Qwen3-Embedding-4B, Zola, Obsidian
Mood: Building.
Upcoming: ClawCon Nashville — Feb 17, co-hosting with Dieter Scholtyssek. Duluth building trip. Taxes.

This is a /now page. It describes what I'm doing right now, not my life story.