I build tools I want to use. Firmware, hardware, automation, infrastructure — and increasingly, full products built through spec-driven AI development. Everything here is built in public and used in daily workflows.
An extension for Pi that turns AI-assisted work into a deliberate loop: scout a codebase, agree on an interactive plan, then build and validate against the resulting spec. Built around a simple belief: the interface with AI is a spec, not a prompt.
View on GitHubA Docker-based sandbox that runs AI coding agents with no git, no SSH, and access only to the directories they need. Built because permission prompts aren't real protection — you end up accepting everything by reflex, and that's not a decision, it's fatigue.pi and opencodeship built-in; adding any other agent takes editing three templates and running a single command.
View on GitHubSpec-Driven Development with AI
An ongoing research practice: building complete fullstack applications through precise specs fed to AI — sometimes working across unfamiliar languages or stacks. A recent example: a full Python web app for managing a music collection, with CSV import, Last.fm API integration, and a custom scraper — built across two evenings with zero prior Python knowledge. The methodology is the experiment. Some projects are public on GitHub.
View on GitHubOpen-sourced PCB + QMK firmware for a 34-key split keyboard and a 12-key dual-encoder pad used for Figma and NeoVim shortcuts.
View on GitHubKeyboard Layer Automation
Open-sourced Node.js utility to switch keyboard layers programmatically based on active workflows.
View on GitHubTelescope Docker Commands Picker
NeoVim extension for Telescope that provides a fast command picker for common Docker workflows inside the editor.
View on GitHubmacOS Calendar Events Binary
Swift script compiled into a binary that displays upcoming Calendar events, designed for SketchyBar integration and reusable elsewhere. Uses EventKit to avoid common macOS TCC permission issues seen in other tools.
View on GitHubOngoing electronics builds, including Arduino programming and DIY HiFi audio experiments.
Ansible-orchestrated Docker services, VLAN-aware firewall networking, and external access via reverse proxy and VPN. Home Assistant dashboards in Grafana powered by sensor data and syslogs.