Aberdeen, UK. Ex-Apple Genius, ten years turning complicated technology into something a person could actually use. I do the same thing now with AI agents.

I run Machines Do It Better, an early-stage solo practice building agent tooling around privacy, data sovereignty and customisation. No clients yet — it is the bet I am making, and I would rather say so than imply otherwise.

Local inference on my own hardware: llama.cpp serving a quantised open-weights model from an RTX 5090, wired into Hermes Agent as an endpoint.

Shipped

• memento-mori — a life-expectancy perspective app, live at memento.technoir.cloud, deployed by me with Docker and Traefik behind a release gate that refuses to publish on red
• hermes-pkm-toolkit — an MCP server for local Markdown vaults, with a filesystem-safety layer and 34 tests covering path traversal
• mental-models-toolkit — thirty machine-readable decision models, packaged for agents
• aberdeen-daily — a nightly local newspaper agent, forked from vaelkeep/hermes-paper-agent and adapted to my city
• hermes-agent-site — an unofficial landing page for Hermes Agent, 45 unit tests, no build step
• javier-ponz-site — portfolio, CV and a case study, source included

Currently learning, kept separate from the above on purpose: vLLM, Proxmox, and a self-hosted homelab.

I write what a thing actually does, not what it aspires to do. If a repo says "MVP", it means MVP.