unit-of-trust
A scrollytelling essay arguing that agent decision legibility should be a first-class layer of the agent stack, not an afterthought.
Next.js/React/TypeScript/Motiondemo ↗
/ prototypes + experiments
Harvey's Lab.
Solo projects I've been building with AI tools — some made in a single sitting, others over a week or two. Mostly Claude Code. I try stuff out and keep iterating on what sticks — little tools I actually use, prototypes still in motion, patterns I'm figuring out as I go.
● Fresh
Also in the lab
↓ the rest, newest first
local tutor
Fully local voice tutor for Mandarin and French — speak, it listens, replies, and talks back. Nothing leaves the machine.
Claude Code/FastAPI/Reactgithub ↗
observe-ui
Decision-legibility surface for autonomous coding agents (live cockpit + retrospective blame).
TypeScript/JavaScript/Pythongithub ↗
Recommend Arena
Twelve recommendation-system designs benchmarked head-to-head on the same dataset and ground-truth queries.
Python/ChromaDB/NetworkXgithub ↗
AI Pattern Testing
A living catalog of AI UX patterns — agent orchestration, tool-call visualization, adaptive cards and UI.
Next.js/TypeScript/Claude Codedemo ↗
Journ
All-local, encrypted, voice-first journaling for iOS and macOS from a single SwiftUI codebase.
Swift/SwiftUI/SwiftDatagithub ↗
Anagnori
Self-guided iOS app for exposure therapy, built around experiments rather than sessions.
SwiftUI/SwiftData/Next.jsdemo ↗
plytools
CLI for using Claude Code to brainstorm and track job applications.
Claude Code/Node.js/TypeScriptgithub ↗
RunTools
Browser-based running utilities — pace converter, heart rate zones, race predictor. Everything runs client-side.
Next.js/TypeScript/Tailwinddemo ↗