Work

Small bets under AvidX. Live ones first. The retired one stays because the reason it died is the useful part.

  • Lellow

    Live · iPad / Web

    Offline, ad-free creative play and games for kids

    Vite · React · TypeScript · IndexedDB · Supabase sync · Renderlellow.app ↗
    • Fully offline PWA, tested in airplane mode
    • 64 coloring pages, freehand and tap-to-fill
    • Cloud backup; kids never see a login
    • No ads, no tracking, no in-app purchases
    Problem
    Kids' apps are either ad-funded attention traps or subscription walls, and most of them stop working the moment the wifi does. I wanted something my own kids could open on an iPad in airplane mode: no account, no ads, no upsell, no network.
    Strategy
    A progressive web app that installs to the home screen and runs entirely offline. Nine activities: freehand coloring and tap-to-fill across 64 line-art pictures, a saved-art gallery, themed word hunts, a math game built on drag-to-feed, a platformer, a plane-dodger, a memory match. Every drawing is a keepsake, so a parent can sign in once and the whole family's art and scores sync across devices. The cast is named after my kids' own words for things: Lellow, Bumbo, Confetti Monster, Aircoo, Gordy.
    The bet
    Built from zero in one evening as an experiment in how fast an idea can reach a real device, then kept because my kids actually use it. The bet is that a calm, offline, ad-free option with a PBS-grade ethos is rare enough that parents will pass it to other parents. Four families are testing it now.
  • Watch Me

    Live · Web / PWA

    A movie and TV watchlist that picks what you watch tonight

    Next.js · Prisma · Postgres · Supabase auth · Claude · Stripe · Resend · Rendergowatchme.app ↗
    • One AI pick a night, with the reason
    • TMDB-backed list for movies and TV
    • Pro tier on Stripe, mobile-first PWA
    • The first AvidX product, still shipping
    Problem
    Streaming platforms recommend what's good for their catalog, not your taste. Your real watchlist is scattered across texts, screenshots, and half-remembered conversations, and the nightly "what should we watch" still burns twenty minutes.
    Strategy
    Own the list first: one place you control, independent of any platform, with full movie and TV metadata from TMDB. Then make the app open on a decision, not a library. Tonight's pick is chosen by Claude from what you've saved, what you've already finished, and the time of day, with a reason attached. Free gets the daily pick. Pro gets to steer it: time, company, energy, not tonight, pick again, alerts when a show you track drops a new season.
    The bet
    People frustrated with platform recommendations will keep a personal list if the payoff is a faster decision. This is the first AvidX product built to charge money, and the one proving the whole paid pattern end to end: Stripe, gating, legal, support, the lot.
  • Planfinity

    Live · Web

    Gridfinity drawer planner for 3D printing

    Next.js · TypeScript · Tailwind · Vitest · Render staticplanfinity.app ↗
    • Drawer dimensions to baseplate plan in one step
    • Rotation, undo/redo, random fill
    • Print summary plus JSON save and load
    • Client-side, no account, works on a phone
    Problem
    The bottleneck in Gridfinity organization isn't the printing. It's the planning. Measure the drawer, work out how many grid units fit, figure out which baseplates to print within a 5×5 bed, decide which bins go where, then hope you got it right before committing hours of filament.
    Strategy
    A single client-side tool. Enter drawer dimensions and get the grid and baseplate split instantly. Place bin footprints on a live grid with rotation, undo, and random fill, then export a print summary and a JSON layout you can reload later. No account, no backend. The whole thing runs in the browser because it doesn't need to be more than that.
    The bet
    Zero-friction tools win in hobbyist communities. No signup means no barrier: someone can share a link and a new user is productive in thirty seconds. The space has since filled with heavier tools that do CAD export and layout optimizers. Planfinity's edge is the fastest path from drawer dimensions to print list.
  • Mail Zero

    Retired · Web

    Gmail inbox analyzer

    Next.js · Claude Haiku · Umami · Render
    • Streaming parser, flat memory at any size
    • Attention model ranks what you actually open
    • Claude-categorized newsletter audit
    • No OAuth, nothing stored
    Problem
    Gmail has no view of who fills your inbox. Most inbox advice is about processing faster, but the real lever is inflow. I wanted to see, with numbers, which senders and newsletters to cut.
    What it did
    Export a Google Takeout archive, upload it, get a full dashboard: inbox health score, email debt, top senders, unsubscribe candidates ranked by a client-side attention model, and a newsletter audit that used Claude to categorize every subscription and flag the ones you never open. Privacy-first by design: no OAuth, no email content stored, nothing kept after the session.
    What happened
    Shipped, worked, and then the ground moved. Once an LLM could plug straight into Gmail and triage an inbox interactively, an upload-a-dashboard product had no reason to exist. Wound down August 2026. If it comes back, it comes back as an agent, not an app.