things I wanted to exist, built at weird hours in san francisco.
A memory that travels with you, across every AI tool you use.
One portable memory for every AI tool you use. Lodis Local is open-source and runs on your own machine — fully readable, fully yours, free forever. Lodis Cloud is the hosted version for syncing across devices — request access to try it.
Your iMessage and WhatsApp, connected to Claude Desktop.
A signed macOS app that connects your iMessage and WhatsApp to Claude Desktop. Once Claude can read your threads, a lot opens up — catching up on what you missed, digging out the detail someone texted last week, drafting the reply you keep putting off. By default nothing sends without you: replies are staged for approval with a hold-to-send button in the menu bar, and that gate is configurable. Built for myself first, but I'm just as excited about what it unlocks for everyone else.
Your sitters, texted in order, until someone says yes.
Book, confirm, and pay the sitters you already trust — entirely over text. No new app for anyone to download, no accounts for sitters to forget. Launching in public beta soon; get on the list and I'll reach out when it's your turn.
Can I build a health coach that leverages all my data without another subscription or wearable?
An agentic health system built from the ground up, starting with how it thinks, not how it looks. The core is a Claude Cowork project with access to my full medical history, lab results, and clinical notes stored as local files. A lightweight iOS app (forked from Gauge) syncs HealthKit and Peloton data to my Mac continuously.
The interesting question isn't the UI. It's the agent design: what should a system know to monitor health meaningfully? What should it surface proactively? When should it ask vs. act? Building this for myself first, deliberately and incrementally, as a way to explore those questions with real data.
An AI coach that actually knows your training.
Gauge tracks your workouts, monitors recovery signals, and adapts your program in real time. The AI coach sees your PRs, readiness, and history — then adjusts volume, intensity, and exercise selection so every session fits how you're showing up that day.
Why it's paused: I built the full app before the core agent system was solid — rebuilding leaner now and planning to grow back toward this vision.
A pictorial routine for toddlers (and the parents running their days).
One of the most amazing things about toddlers is how much they can do and manage on their own. Routines help immensely. While his iPad gathers dust between plane flights, I thought I'd put it to use: a pictorial routine calendar so that he can see and interact with his day, laid out in friendly cards — wake up, play, breakfast, get dressed — with educational mini-games mixed in.
Why it's paused: the interaction pattern doesn't fit a toddler yet — he can't drive it on his own, and if it's parent-in-the-loop anyway, an analog form factor probably serves better than a screen.
Experimenting with building an autonomous code development platform.
An orchestrator that ended up too close to a clone of Claude Code itself. It might be worth repurposing into a DevOps/bug fixing tool, but my vision of going from me writing a PRD to Agent Forge writing finished software didn't work reliably. It also became expensive to run, consuming a ton of tokens via API vs. my (presumably subsidized) Claude subscription. On the shelf for now, but it earned solid learnings about structuring data and workflows for autonomous agentic work.
Retired as a discrete product; what survives is a set of personal MCP tools that beat any out-of-the-box connector. The LLM is the frontend.
Everyone wants a personal assistant, right?! This was my first playground. The further I got with it the more I realized:
So many lessons learned here have spawned into projects around this portfolio.
I'm slowly still building towards this mission. The first step was keeping the useful tools and putting them into an MCP that Claude can access. For example, Gmail's scope is too limited with the out of the box connectors, so I gave Claude additional scope through this MCP tool.
Set-and-forget restaurant reservations — catch the drop, pounce on the cancellation.
My first real build: a bot that would grab hard-to-get reservations the instant they were released and snipe cancellations as they opened up. It turned into a cat-and-mouse game with bot detection and never got reliable enough to trust. The punchline came later — a one-off scheduled Cowork task driving an authenticated Chrome session actually does the job, as long as you're diligent about setting it up each time. Rez Sniper was trying to automate away exactly that diligence; it just couldn't clear the reliability bar. Killed it, but it's the project that taught me to build.
A live read on your Claude and Codex burn, so you don't hit the wall mid-session.
A menu-bar dashboard that puts your Claude and Codex usage in one place — session and weekly limits across both, when each one resets, your projected monthly bill, and what that same usage would've cost at API list prices. Forked from Claude God and bent toward the question I actually had every day: how much have I got left before I hit the wall? The name's a play on "token maxxing" — run hot, just don't run out.
My dad wanted to rate everything he watches. Two hours in Codex later, he had it.
What began as a simple log grew a watchlist, a recommender, and viewing stats nobody asked for. The real exercise was velocity — how complete a build I could get out of one tight iteration loop in a single sitting, empty repo to live URL. Complete enough that it stuck.
