A complete list of public work, plus experiments that are interesting but not flagship-ready.
My favorite projects reduce tooling overhead so agents feel instant. When startup costs cross the "noticeable" threshold, workflows become frustrating fast.
See: iMessage Gateway CLI, Gmail/Calendar daemon paths.
Full-stack MMA fighter database with 76,000+ fighter profiles, 270,000+ fights, and real-time stats. A Pokedex-style experience for combat sports enthusiasts.
AI-native platform for financial research providing LLM-ready SEC filing data with semantic search, RAG chat, and structured table extraction.
A searchable archive of court documents with OCR text extraction and full-text search capabilities. Makes publicly released legal documents accessible and navigable.
Interactive periodic table with animated electron orbital visualizations. AI-powered exploration of atomic structure and quantum mechanics concepts.
Historical weather context for any location. Compare today's conditions against 100+ years of NOAA climate data with interactive visualizations.
Wolfie's portfolio website built with React, TypeScript, and Framer Motion. Features WebGL shaders, fluid animations, and a dark terminal-inspired aesthetic.
MCP server enabling AI-assisted music production in Ableton Live. 50+ tools for MIDI manipulation, automation, groove control, and arrangement—all controllable through natural language.
Provenance for AI-written code: prompt → patch → verification → merge. “Git for agent actions”.
The fastest iMessage integration for Claude Code. CLI + daemon design avoids MCP overhead for "instant" UX.
A shareable inventory + analytics layer for your Claude Code setup, with an interactive TUI dashboard.
A personal operating system: projects, tasks, notes, and integrations—driven through an agent-first CLI.
Turn raw LLM chat exports into structured datasets and research-friendly artifacts.
Analyze your full Spotify listening history locally in the browser (not just 90 days). Privacy-first, no server required.
A Python library for flight search workflows with clean abstractions and reproducible results.
Early-stage experiment exploring social + ML ideas.
Daemon-based architecture for AI agent tools. 19x faster than MCP stdio with sub-10ms latency via persistent UNIX socket daemons.
Precinct-level map of San Francisco elections built with deck.gl and MapLibre. Recolor ~514 precincts by leading candidate across contests, with a correlator mode comparing scraped civic data against certified vote share.
A unified 3D civic-stress map of San Francisco - poop, crime, fire/EMS, 311, and traffic crashes rendered as one multi-source hex stack with a rolling 91-day playback window and a diverging-delta compare mode.
A real-time urban-intelligence map of San Francisco: crime, fire, building permits, landmarks, heritage sites, and 311 service requests layered over 3D buildings with deck.gl.
A unified 3D map of San Francisco civic event streams - 311 waste and needle reports, SFPD crime, SFFD fire calls, general 311 cases, and SFMTA traffic incidents - with every rendering mode exposed as a runtime toggle.
A global country-statistics dashboard: compare nations across 60+ datasets with choropleth maps, correlations, trends, and composite indices, all with shareable URL state and validated data.
Google Maps for biodiversity: pan and zoom the entire phylogenetic tree of life, drilling from Domain to Species with iNaturalist photos, range maps, and Wikipedia summaries at each leaf.
An interactive 3D visualization of ~300,000 human and animal waste reports filed with SF 311 between 2008 and 2024, rendered as a deck.gl hex map.