20 projects

All the
good stuff.

Agent systems

5 projects

Active, MIT

Nexus Office

Agents that keep working without you need a place where you can see and answer them.

Public MIT repo with live Mac and phone clients.

Active, on the Claude plugin marketplace

KERNEL

Claude Code kept starting over, so I built memory, rules, and receipts around it.

360 commits since January 2026, on the Claude plugin marketplace.

Active, MIT

the-agent-library

Prompt collections kept rotting, so I turned repeated workflows into portable skills.

39 skills, each extracted from repeated real-world use.

Published on PyPI, MIT

agentmailkit

A scheduled email should look the same every day, even when a model writes the words.

36 offline, deterministic tests. Dry runs cannot send, enforced in code.

Live

Substrate

I wanted to see what an unattended creative pipeline would become if it kept going.

425 pieces and counting, generated daily without supervision.

Memory and context

4 projects

Personal macOS tool

Hearth

The notch could be a quiet home for the context I keep reaching for.

Runs locally on macOS, with deterministic tests for transcript parsing and conversation status.

Shipped to production, 2025 to 2026

HeyContext

Context kept disappearing, so we built a workspace around memory.

Went live with hundreds of users within a month, no ad spend.

Integrated into HeyContext

HeyContent

Creators had context everywhere and nowhere, so we tried to bring it into one place.

5+ platforms integrated with real-time sync. The memory layer survived into the next company.

Published on PyPI

metabrain

Memory gets noisy unless it has to prove itself.

Published on PyPI. Zero dependencies by design.

Evals and verification

5 projects

Active

site-spec

A website can look finished while everything machines need is quietly broken.

Live-URL and local audits, JSON output, CI-ready exit codes, deterministic site builds.

Published on npm, MIT

renderstate

A screen should be inspectable in every meaningful state before the whole app can run.

Published on npm with multi-viewport capture.

Active

llm-bench

Leaderboards were not answering my questions, so I made tests that did.

21 tests, graded by programmatic verifiers.

Bootstrap loop shipped

model-familiarity-engine

I wanted to know what a model had earned across a working relationship, not where it ranked.

Replay-bootstrap loop shipped: redaction, replay, observed model cards.

Research, negative results preserved

latent-diagnostics

Correct answers were not enough. I wanted to know whether the model computed something real.

The failed hypothesis is documented next to the confirmed one.

Products

6 projects

Live on the App Store

ModelMind

AI education felt backwards, so I built the course I wanted people to have first.

Live on the App Store for iPhone, iPad, and Mac.

Live on the App Store

Paper Rooms

Safari tabs ate my research life, so I built a library.

Built and shipped solo in under a week.

Live

our4cuts

An iPad and a browser are all you need for a photo booth.

Live and used at real events.

Live on the App Store

Civic Forges

I wanted a city builder where running the city and living a life are the same game.

On the App Store and playable in the browser, with persistent saves and 36 automated tests.

Live on the Chrome Web Store

Not Recommended

YouTube should begin with a question, not an infinite feed.

No accounts, analytics, ads, or remote executable code.

TestFlight, 2024 to 2025

Brink Mind

Mental health tools ignored the body, so Brink Mind brought heart data into the room.

Reached TestFlight with working voice, biometrics, and on-device inference.