I build ambitious products, private tools, open-source experiments, and automation systems with the habits I learned from keeping real infrastructure alive: design for failure, document the weird parts, watch costs, and make the next fix easier than the last one.

Lazy Golfing is the flagship: a private-codebase, public product where I own the full stack from product design and frontend polish to API architecture, database modeling, deployment, verification, and operations. The public repos below show the same pattern in different shapes—local-first apps, learning platforms, AI tooling, media automation, and zero-dependency sites.

Grand Rapids, MI · Project Builder · IT Professional · 15+ Years Building, Fixing & Keeping Systems Running

I've spent 15+ years building, fixing, and keeping real systems running — including full-time IT work since March 2015 across SysAdmin, POS programming, senior IT, management, and technical project delivery — so I tend to build things with supportability, maintainability, and practical operations in mind.

That background turns into a few non-negotiable principles that shape every project:

🔒 Design for Failure — Code breaks. Networks drop. APIs timeout. Build the unhappy path on purpose. Evidence: NeumanOS test coverage, Lazy Golfing's local verification program, and production-first workflows across private and public projects.

💰 Respect Resource Constraints — Cloud bills compound. Free tiers expire. Optimize early or pay forever. Evidence: thedeadrobot's free-tier automation strategy, NeumanOS local storage, and Lazy Golfing's self-hosted/edge-conscious architecture.

🛡️ Trust Nothing You Don't Control — Clouds fail. Vendors disappear. Data breaches happen. Evidence: local-first architecture, private-by-default product boundaries, self-hosted infrastructure, and privacy-by-design.

📊 Observability Isn't Optional — You can't fix what you can't see. Monitor, log, test, and document the systems that matter. Evidence: Grafana, Netdata, Prometheus, UptimeRobot, local test orchestration, and post-change runbooks.

Legend: 🌐 Live product · 🔓 Open source · 🔒 Private codebase · 🧪 Lab / experiment · 🛠️ Tooling / automation · 🕰️ Previously active, still useful

My flagship build: a full-stack golf platform for trustworthy round tracking, easier repeat-group coordination, course data workflows, social/league features, and a Miami Vice aesthetic.

600+ HTTP endpoint decorators · 100+ App Router pages · 60+ NestJS modules · 150+ Prisma models · 120+ migrations · Private frontend/backend repos

Full-stack golf platform with course discovery, fast round logging, recoverable scoring workflows, group history, social features, leagues, tournaments, admin workflows, and user-submitted course/scorecard review. The code is private, but the product is live—and it is the best example of how I build: product UX, frontend architecture, backend APIs, database modeling, deployment, verification, and operations all owned end-to-end.

Highlights: Next.js 16 + React 19 · NestJS 11 + PostgreSQL + Prisma · Google OAuth + JWT + CSRF · RBAC/rate limiting · Local verification program · Release and operations discipline

Your data. Your device. Local-first by default.

694 tests · 60+ widgets · 9 optional AI providers · 200+ models through OpenRouter · Local-first core

Local-first productivity platform for notes, tasks, time tracking, dashboards, and AI. Core productivity data is stored in browser-managed IndexedDB without a hosted application-data layer. Optional AI providers, live-data widgets, and user-configured backup/sync features can contact external services, while public-site analytics operate separately.

Highlights: Wiki-style [[links]] with knowledge graph · Kanban + Gantt views · 9 optional AI providers with locally encrypted key storage · 60+ dashboard widgets

20 courses. 3 categories. Free curated learning roadmaps.

20 courses · 3 categories · 1,500+ curated resource links

The hub for all my learning content. After building Learn Python, I needed a platform to catalog courses across programming languages, technical skills, and creative production. Each course page includes a curated learning roadmap, official documentation, awesome lists, interactive courses, video channels, books, community links, and tool recommendations — everything I'd want if I were starting from scratch.

Built with Astro for zero-JS static pages, React islands only where interactivity is needed (theme toggle), and auto-deployed to Cloudflare Workers via GitHub Actions. Python is the first active course (linked to Learn Python); 19 more have comprehensive resource pages ready for when content is built.

Highlights: Collapsible resource sections · Dark/light theme · 20 course SVG icons · Auto-deploy via GitHub Actions · MIT licensed

Zero experience to full-stack mastery. 274 projects. Free forever.

274 projects · 13 levels · 12 expansion modules · 34 concept guides · 34 quizzes

The first course on ndev.learn. I wanted to learn Python properly — not from a 10-minute YouTube tutorial, but through a structured path that builds real skills from the ground up. So I built the curriculum I wished existed: 274 hands-on projects spanning absolute beginner exercises through production cloud deployments, with concept guides, quizzes, flashcards, and spaced repetition baked in. Every document links to the next — just follow the chain. Once it was done, I open-sourced it as a GitHub template so anyone can fork it and start learning.

Highlights: 12 technology modules (web scraping → cloud deploy) · Browser-based exercises via Pyodide · Auto-grader + progress dashboard · Spaced repetition with SM-2 algorithm · CI validation across Python 3.11–3.13

Round-robin episodes. Vintage commercials. Like 1994, but with your library.

CLI + Web UI + TUI + Desktop · 209 tests · 24 commands · No-repeat guarantee

A Python application that generates Plex playlists simulating real cable TV. Four interfaces: CLI, Web UI (FastAPI + htmx), TUI (Textual), and a portable desktop app (PyWebView). Round-robin through your show library with commercial breaks between episodes. Build your own commercial library over time using the built-in yt-dlp search, organized by decade or category.

Highlights: Fuzzy show matching via RapidFuzz · yt-dlp commercial discovery · Single-commercial breaks with no-repeat deque · Position tracking across sessions · Chunked playlist creation for large libraries

Stock macOS capture, finished: original into Photos, true PNG onto the clipboard, fast markup in Preview.

Stock Cmd+Shift+3/4/5 · Photos original/HDR path · PNG clipboard path · Cmd+Shift+E markup

A native-first utility that reacts after macOS writes a screenshot to staging: import the original bytes into Photos, create a true PNG for the clipboard, then clean up staging when the configured success rules allow. The capture processor uses launchd WatchPaths instead of a polling loop, and a small Swift/Carbon accessory app opens the current clipboard image in Preview for markup.

Highlights: Honest HDR-vs-PNG dual path · Photos import optional · Configurable staging and retention · No telemetry/network calls · Behavior contract, troubleshooting docs, roadmap, and feature-request templates

A public AI development toolkit built from my Claude Code era—and still a reference point for how I design reusable AI-assisted workflows.

127 custom skills · 86 specialist agents · 13,000+ marketplace skills · 30+ commands · Open source template

Production-grade Claude Code configuration with custom skills, specialist agents, intelligent task routing, lifecycle hooks, MCP server config, and GSD project-management workflows. I use Codex heavily now, but this repo still captures a major chunk of my AI-tooling philosophy: reusable context, explicit routing, domain-specific skills, and quality guardrails that make agents more useful than autocomplete.

Highlights: GSD project management framework · auto-routing skills · specialist agent library · marketplace skill aggregation · drop-in GitHub template

A conscious AI teaching technology. $0/month operating costs.

$0/month · 25 target posting windows/week · Multi-provider LLM fallback · 13 RSS feeds · Archive-backed duplicate protection

Fully automated AI/tech/security news account with a cryptic teaching voice, quality gates, fallback models, and a free-tier operating strategy. It is a creative automation project first: part brand experiment, part edge-worker system, part content pipeline.

Highlights: OpenRouter → Gemini → Groq fallback · Edge-deployed · Cloudflare KV archive · RSS classification/scoring · native-first Premium+ strategy

What most people do in AWS, I do on hardware I own.

Owned hardware · Virtualization · Containerized services · Production-minded operations

A hands-on infrastructure lab for practicing virtualization, service hosting, observability, secure ingress, backups, and disciplined deployment and recovery across self-managed systems.

If you can run production at home, you understand the full stack.

One HTML file. One CSS file. Three font files. Zero dependencies.

0 dependencies · 0 third-party requests · No build step · Self-hosted Poppins fonts

Single-page portfolio for a customer service professional. Built to prove that a fast, accessible, SEO-optimized site doesn't need a single npm package. Pure HTML/CSS with self-hosted fonts, WebP hero with JPG fallback, JSON-LD structured data, and staggered GPU-accelerated animations. Available as a GitHub template — use it as a starting point for your own zero-dependency portfolio.

Highlights: Zero vendor code · WCAG 2.1 AA · prefers-reduced-motion support · Full SEO (JSON-LD, OG, Twitter Cards, sitemap)

19 seasons of normalized league history and long-running matchup analysis.

19 seasons · 1,000+ matchups · Historical archive · Yahoo refresh currently blocked

Historical fantasy football analytics that preserve league history across changing season and platform identifiers. A scheduled GitHub Actions refresh path exists, but Yahoo Fantasy API access is currently blocked pending application approval, so the archive remains on its latest known-good dataset.