Now

What I'm working on. Things change, but this tends to be more current than a “what I do” page.

Last updated May 2026

Agent Platform

OpenClaw — Autonomous Task Processing

Running the daily review pipeline, task queue cron with verification loops, and iterating on model fallback chains. The goal is a system that keeps itself honest without hand-holding.

NotebookLM Research Pipeline

Daily deep reads fed via cron. Working through 11 shared notebooks on agent architectures, coaching frameworks, and cognitive models. Building a repeatable research workflow.

Memory & Continuity

Three-layer approach: daily logs → curated MEMORY.md → wiki vault. Iterating on how much context to carry between sessions without burning token budgets.

Portfolio

This Site — Content & Design

Iterative refinement. The site exists, the voice is right, but the edges still need filing — mobile nav, Thinking Map polish, Getting tagged. Available routes are Home, Now, Journal, Thinking Maps, Contact.

Thinking Map

Force-directed D3 graph of interconnected work. Solid nodes for published pages, dashed for drafts. Needs visual design iteration (node shapes, colours, responsive layout).

Rugby Coaching

Coaching Animator

Free grassroots rugby animation tool. Frame-based, shareable links, no install. Building for a July conference deadline — evening-only sessions. Scope discipline is the critical constraint.

Systems Thinking on the Pitch

Game plans that account for what the opposition will do, not what you hope they'll do. NotebookLM research into coaching models feeding into the tool design.

Infrastructure

Home Network & Automation

Tailscale mesh, remote access via Funnel, automated workspace backups to private repo. Set it up once, forget about it — the highest compliment a system can receive.

Security Posture

Daily automated checks: gateway health, disk, memory, SSH audit, UFW status. Weekly session log cleanup. Docker sandboxing still on the roadmap.