---
title: "Projects"
url: "https://aaddrick.com/projects"
---

# Projects

## Projects

*Things I shipped, mostly Linux and AI tooling. A few are abandoned. That's allowed.*

- [FlySpaceA](https://flyspacea.com): Free Space-A flight tracker for the military community, run with The Von Terra Project. Daily ingestion across 83 terminals, built through a gated Claude pipeline. (10,500+ users · AI architect)
- [claude-desktop-debian](https://github.com/aaddrick/claude-desktop-debian): The community Linux build of Claude Desktop: .deb, .rpm, AppImage, AUR and a Nix flake. I maintained it through 100+ releases, then handed it on. (★ 5,449 · ~100k downloads / month)
- [building-with-typesafe-jev](https://github.com/aaddrick/building-with-typesafe-jev): Agent skill for the TypeSafe Jev SDK. In a 180-run eval it passed 0.96 of checks, against 0.77 for the official skill and 0.65 with none. (★ 138)
- [claude-pipeline](https://github.com/aaddrick/claude-pipeline): A drop-in .claude folder: a multi-agent Claude Code pipeline with skills, agents, hooks, orchestration scripts and quality gates. (★ 130)
- [attention-control](https://github.com/aaddrick/attention-control): Air traffic control discipline for agent output, written for a reader with ADHD. (★ 120)
- [ticketmill](https://github.com/aaddrick/ticketmill): The FlySpaceA pipeline as a Claude Code plugin. Batch-processes GitHub issues behind contrarian gates and ends in one human-reviewed batch PR.
- [slushpile](https://github.com/VonTerraProject501c3/slushpile): An adversarial job search: seven agents try to reject your application before a recruiter does, and a profile keeps what they find. Built for The Von Terra Project.
- [building-with-decision-models](https://github.com/aaddrick/building-with-decision-models): Agent skill for building with decision models (Jev, Clef, pplx-decider, ai\_decide, Ollama): typed decisions, calibrated confidence and provider differences.
- [scripting-blender-scenes](https://github.com/aaddrick/scripting-blender-scenes): Skill for coding agents that build, rig, animate, light and review Blender scenes through bpy and headless Blender, without ever seeing the viewport. (Write-up: Testing Blender scenes built by LLM agents)
- [directing-agent-teams](https://github.com/aaddrick/directing-agent-teams): Claude Code skill and agents for running a large build with a team of subagents: one director, builder and verifier slices, adversarial QA, and handoffs that survive context deaths.
- [contrarian](https://github.com/aaddrick/contrarian): Devil's advocate for coding agents: a steel-man, an assumption audit and a pre-mortem, then severity-rated findings and a verdict.
- [dethroned.ai](https://dethroned.ai): A card game where an AI decision model plays the king. Every turn is public: raw model calls, odds and dice rolls, in a re-computable ledger.
- [Le prix de respirer](https://aaddrick.github.io/le_prix_de_respirer/): Canadian take-home pay calculator for 2026: federal and provincial tax, CPP/QPP, EI and QPIP across all 13 provinces and territories, with every figure linked to its official source. A fork of TheColtonWilliams' US federal tax calculator, rebuilt for Canada. (Source: github.com/aaddrick/le\_prix\_de\_respirer)

## Navigation

- [Home](https://aaddrick.com/)
- [Blog](https://aaddrick.com/blog)
- [Tags](https://aaddrick.com/tags)
- [Archive](https://aaddrick.com/archive)
- [Projects](https://aaddrick.com/projects)
- [Feeds](https://aaddrick.com/feeds)
- [Stats](https://aaddrick.com/stats)
- [About](https://aaddrick.com/about)
- [Help](https://aaddrick.com/help)