some of my favorite projects

A semantic layer over my life data. Each app (Recipe Hub, Music Hub, Wealth Hub, etc.) owns its data in portable formats: ChordPro for music, JSON for the rest. All data is backed up nightly. Life Hub ties it together into one model of people, things, places, events and decisions, with references back to the apps that hold the detail.

All of it is exposed through a single hosted MCP connector with OAuth, so any AI agent from any provider (ChatGPT, Claude, whatever comes next) can read and update my life data. The data stays mine and portable; the assistants are interchangeable.

Every record keeps its provenance: whether I said it, an app recorded it, or an AI inferred it, with a confidence and who asserted it. My own words are stored verbatim as the source. Every change is a new version I can see and undo.

• Any AI agent ChatGPT, Claude or whatever comes next, through one MCP connector: OAuth sign-in, per-connection limits, an audit log, and a fixed set of tools (capture, search, get, retract, songs).
• Life Hub Records (entities, relationships, events, decisions, sources), each with provenance (stated, recorded or inferred; confidence; who asserted it) and full, undoable history.
• The apps Recipe Hub and Music Hub hold the detail; Life Hub records link to them. Wealth Hub is planned.
• Backups Nightly snapshots of every hub.

• MCP
• OAuth
• Cloud Functions
• Firestore
• append-only versions

Recipes with versions and baker's percentages, sessions with a step-by-step walk-through, ratings and push reminders. It started as a sourdough log and grew. Most entries arrive through a chat with ChatGPT or Claude; the app is where I check, cook and rate.

• vanilla JS PWA
• Firestore
• web push
• MCP connector
• Playwright

Songs are stored as ChordPro, an open text format. It imports from the web or a photo, transposes, and keeps every change with before and after, so an assistant can fix a chord and I can undo it.

• ChordPro
• vanilla JS PWA
• Firestore
• MCP tools via Life Hub

Ziggy talks, remembers what you tell it, says what it sees, drives to named rooms, carries messages between them, and has a locked-down mode for playing on the floor with a toddler. The talking half comes first; the wheels are mostly a robotics-learning project.

cloudfrontier modelconversation and reflection

Jetson Orin Nanothe mind on boardmemory, identity, senses, voice · ROS 2

ESP32-S3the reflexesreal-time motor safety

• ROS 2
• Jetson Orin Nano
• ESP32-S3
• Python
• PlatformIO

Drag, throw, drop and splash colorful shapes and animals that bump into each other and float in water. Tap something and it says its name out loud and shows the word; tap empty space and a new one appears, named the same way. Preset scenes (farm, desert, beach, under the sea), bold high-contrast shapes, and no goals or failure: pure cause and effect, with a few new words along the way.

• Godot 4
• GDScript
• iOS first

A solo project from the ground up: a place for my wife and me to record and share our love for Santa Barbara. Originally Next.js on Vercel, since rebuilt on React Router, with AI-assisted publishing tools and an MCP connector so ChatGPT and Claude can search it.

• React Router
• Supabase
• Fly.io
• Cloudinary
• OpenAI
• MCP

The Sinclair Digital team builds and maintains the web technology for every local news station SBG owns. I helped build a new hybrid mobile app in React Native and maintained the existing web apps.

The interesting challenge was white-labeling: one codebase that serves every station, configured for each one's branding and content.

the KOMO News app (Seattle), one station's build of the white-label app

• React
• React Native
• white-label platform

KOMO News (example station) iOS app Android app

I led the front-end team and worked hands-on across the whole system, designing and building it end to end. We were a scrappy team of six, and we helped grow the startup to ~$30 million in funding and ~50K monthly unique users.

We wrote an enormous amount of code and shipped features at a blazing pace. The demo above walks through the three core apps we built and ran: the customer web app, the mobile apps and the back office.

• React
• Flutter
• Hasura / GraphQL
• PostgreSQL
• Firebase / GCP
• AWS
• Stripe
• Invers telematics
• A/B testing

iOS app Android app