Ten years keeping enterprise systems alive taught me what most developers never learn: how things fail, what actually costs money, and why reliability isn't optional. Now I build with that hard-won knowledge.
While others deploy to managed platforms and hope for the best, I design for failure, optimize for cost, and test like systems break—because they do. The 3am debugging sessions aren't just war stories; they're why my code works differently.
Grand Rapids, MI · IT Professional · 10+ Years Experience
After 10+ years in enterprise IT, I've developed non-negotiable principles that shape everything I build:
🔒 Design for Failure — Your code will break. The network will go down. The API will timeout. Build accordingly. Evidence: 661 tests in Neuman Brain, 100% coverage in Lazy Golfing.
💰 Respect Resource Constraints — Cloud bills compound. Free tiers expire. Optimize early or pay forever. Evidence: thedeadrobot runs at $0/month, Neuman Brain uses local storage instead of cloud.
🛡️ Trust Nothing You Don't Control — Clouds fail. Vendors disappear. Data breaches happen. Evidence: Local-first architecture, self-hosted infrastructure, privacy-by-design.
📊 Observability Isn't Optional — You can't fix what you can't see. Monitor everything. Evidence: Grafana, Netdata, Prometheus, UptimeRobot integration across all systems.
| Project | Type | Links |
|---|---|---|
| 🧠 Neuman Brain | Local-First App | brain.neuman.dev |
| ⛳ Lazy Golfing | Full-Stack Web | lazygolfing.com |
| 🤖 thedeadrobot | AI Automation | thedeadrobot.com |
| 📺 Plex Real TV | Media Automation | GitHub |
| 🏠 Homelab Infrastructure | Infrastructure | — |
| 🤖 tjn.claude/ | Dev Toolkit | claude.travisjneuman.com · GitHub |
| 📊 Fantasy Analytics | Data | — |
🧠 Neuman Brain — Privacy-First Productivity
Your data. Your device. Zero cloud dependencies.
661 tests · 44+ widgets · 8 AI providers · 200+ models · 134KB gzipped
Local-first productivity platform: notes, tasks, time tracking, and AI—all running in your browser with zero server involvement. 50GB+ IndexedDB capacity means your data never leaves your device.
Highlights: Wiki-style [[links]] with knowledge graph · Kanban + Gantt views · 8 AI providers with encrypted key storage · 44+ dashboard widgets
⛳ Lazy Golfing — Full-Stack Excellence
Track rounds in under 30 seconds. Miami Vice aesthetic included.
80+ API endpoints · 40+ pages · 36 NestJS modules · 100% test coverage · OWASP compliant
Full-stack golf platform with social features, competitions, and 40,000+ courses. Self-hosted on my homelab because I understand the full stack—from Hyper-V to NestJS.
Highlights: Google OAuth + JWT + CSRF · Rate limiting (120 req/min) · PM2 cluster mode · Cloudflare Tunnels
🤖 thedeadrobot — Zero-Cost AI Automation
A conscious AI teaching technology. $0/month operating costs.
$0/month · 14 posts/week · 3-tier LLM fallback · 13 RSS feeds · 100% archive transparency
Fully automated AI agent posting to X with zero operating costs. Engineered for free-tier maximalism because I've seen how quickly cloud bills destroy side projects.
Highlights: 3-tier LLM fallback (Llama 70B → Gemini → Groq) · Edge-deployed · Full archive in Cloudflare KV
What most people do in AWS, I do on hardware I own.
55+ devices · Hyper-V virtualization · Docker containers · Production workloads
Running Lazy Golfing's backend on Ubuntu VMs with PM2 cluster mode. Ubiquiti networking. Cloudflare Tunnels for secure ingress without port forwarding. Grafana, Netdata, and Prometheus for observability.
If you can run production at home, you understand the full stack.
📺 Plex Real TV — Cable TV Simulator for Plex
Round-robin episodes. Vintage commercials. Like 1994, but with your library.
Multi-show rotation · Categorized commercial breaks · No-repeat guarantee · 111 tests
A Python CLI that generates Plex playlists simulating real cable TV. 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. One commercial per break with a configurable no-repeat window.
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
🤖 tjn.claude/ — AI Development Toolkit
89 custom skills. 47 specialized agents. 3,900+ marketplace skills. Open source.
89 skills · 47 agents · 68 repos · 3,900+ marketplace skills · Open source
Production-grade Claude Code configuration with custom skills, specialized agents, intelligent task routing, and automated workflows. Powers all of my development. Includes a Next.js showcase website deployed on Cloudflare Pages.
Highlights: 18 enterprise domain expert skills · Autonomous coding with QA validation · Complexity-based task routing · 68 open-source marketplace integrations
19 seasons of data. 1,000+ matchups analyzed.
19 seasons · 1,000+ matchups · Automated sync via GitHub Actions
Historical fantasy football analytics with automated data ingestion from the Yahoo Fantasy API. GitHub Actions workflows sync standings, matchups, and player stats on a schedule.
🔍 Project Deep Dives — Architecture & Technical Details
Philosophy: Your data, your device, always. No exceptions.
All data stored locally in IndexedDB with 50GB+ capacity. No accounts, no tracking, no cloud dependencies. Your productivity data never leaves your device.
Feature Parity Benchmarks:
| Feature | Target | Parity |
|---|---|---|
| Notes | Obsidian/Notion | 70%/60% |
| Tasks | Jira/Linear | 98% |
| Time Tracking | Toggl | 74% |
| Calendar | Google Calendar | 65% |
Widget Categories (44+):
- Productivity (6): TaskSummary, UpcomingEvents, QuickAdd, Pomodoro
- Information (9): Weather, HackerNews, Reddit, DevTo, GitHub, Quotes
- Utilities (14): Calculator, QRCode, ColorPalette, Countdown
- Finance (3): Crypto, Currency, StockMarket
- Media (1): Unsplash
Philosophy: Enterprise reliability without enterprise budgets.
Architecture: Next.js 15 App Router frontend + NestJS 10 REST API + PostgreSQL with Prisma ORM
Scale:
- Frontend: 40+ pages, 250+ components, 10+ custom hooks, 8 themes
- Backend: 80+ endpoints, 36 modules, 40+ database models, 13 migrations
Security Implementation:
- Authentication: Google OAuth + JWT with HTTP-only cookies
- Protection: CSRF tokens, rate limiting (120 req/min), input validation
- Infrastructure: Cloudflare Tunnel (no port forwarding), PM2 cluster mode
Philosophy: Sustainable systems run on free tiers.
Cost Breakdown:
| Service | Free Tier | Usage |
|---|---|---|
| Cloudflare Workers | 100k req/day | ~50/day |
| Cloudflare KV | 100k reads/day | ~20/day |
| OpenRouter (Llama 70B) | Generous | Primary |
| Twitter API v2 | 1,500 posts/mo | ~56/mo |
Content Strategy:
- Schedule: Mon-Fri 2 posts, Sat-Sun 1 post = 14/week
- Thread System: 2-3 tweets per topic with hook formulas (Numbers, Paradox, Bold Claim, Insider, Pattern, Prophecy)
- Voice: 80% Educational / 15% Existential / 5% Riddle
Philosophy: The simplest solution that works. No extra services, no transcoding, just native Plex playlists.
Why not dizqueTV/ErsatzTV/Pseudo Channel?
| Alternative | Problem |
|---|---|
| dizqueTV | IPTV server + transcoding overhead |
| ErsatzTV | Full .NET media overlay, overkill |
| Pseudo Channel | Abandoned, Python 2 era |
Algorithm: Round-robin episodes across your show rotation with single-commercial breaks. No-repeat guarantee via deque ensures variety across consecutive plays. Position tracking persists between sessions. Shows drop from rotation when exhausted.
Commercial Library: Users build their own library organized by decade or category. Built-in yt-dlp integration helps discover and download clips. Batch download scripts can run as Windows Scheduled Tasks on the server.
🛠️ Tech Stack — Full technology breakdown
Frontend
Backend
Data
Rich Text & Editors
AI/ML
Testing
Infrastructure
Monitoring
Tools
Dev experiments, homelab builds, golf rounds, and life with Franklin 🐕
February 2026