{"$schema":"https://ejfox.com/schema/page-twin@1.json","kind":"project_index","url":"https://ejfox.com/projects","json_url":"https://ejfox.com/projects.json","generator":"ejfox.com/json-twin@1","data":[{"slug":"projects/MorningRadio","title":"MorningRadio","html":"

Designed to be a healthy \"first screen\" in the morning. News summaries, updates, and enough information about the latest events that you don't feel the need to endlessly doomscroll.

\n

The goal was to repurpose the tik-tok style endless vertical scroll but instead fill it with interesting and useful websites, tools, and news articles (populated through my Pinboard bookmarks)

\n
\n\"Home\n
Home — an audio digest up top, then \"Top Picks\" and a recent feed pulled from my Pinboard bookmarks
\n
\n
\n\"The\n
The morning audio briefing — a generated \"Morning Radio\" segment with scrubber and skip controls
\n
\n
\n\"A\n
A single card in the vertical scroll — full article with source, tags, and summary
\n
\n

I even experimented (unsuccessfully) with creating a home screen widget.

\n
\n\"Home\n
Home screen widget experiment
\n
","metadata":{"title":"MorningRadio","date":"2025-08-10T04:00:00.000Z","modified":"2025-08-13T15:26:19.000Z","url":"https://github.com/ejfox/MorningRadio","tech":["Swift","iOS","Audio Streaming","Mobile UX"],"state":"doing","aiInvolvement":"ai-assisted","tags":["ios","health"],"words":123,"images":4,"imageDetails":{"total":4,"cloudinary":4,"withDimensions":0},"links":4,"codeBlocks":0,"headers":{},"toc":[],"type":"post"}},{"slug":"projects/smallbot-mcp","title":"Smallbot MCP","html":"

Smallweb is a really unique tool to create a filesystem based system for rapidly-deployed deno apps, and it pairs really well with a robot companion.

\n

Nowadays when I have an idea for a website and I just want it to exist, without spending three hours setting up a build pipeline configuring DNS. Smallbot is an incredible tool to defeat that particular form of procrastination—a robot that takes \"make me a site about northeastern trees\" and actually builds the thing.

\n

I made a video with the creator of Smallweb about how easy it is to get setup, and I love how much the project embodies indieweb philosophies

\n

This is a tool built on the MCP (Model Context Protocol) so it can talk to Claude and other AI assistants and give them the tools and documentation to make little Smallweb sites easily, on-the-fly.

\n
\n\"Smallbot\n
Smallbot — a helpful assistant for creating Smallweb applications
\n
\n
\n\"Smallbot\n
Smallbot generating a Smallweb app
\n
","metadata":{"title":"Smallbot MCP","date":"2025-08-02T04:00:00.000Z","modified":"2025-08-13T17:17:13.000Z","url":"https://github.com/ejfox/smallbot-mcp","tech":["MCP","AI","Web Generation","Node.js"],"state":"deployed","aiInvolvement":"ai-enhanced","tags":["ai","web"],"words":156,"images":2,"imageDetails":{"total":2,"cloudinary":2,"withDimensions":0},"links":6,"codeBlocks":0,"headers":{},"toc":[],"type":"post"}},{"slug":"projects/cloudinary-backup-tool","title":"Cloudinary Backup Tool","html":"

I needed to backup thousands of photos from Cloudinary, and couldn't find an easy standalone tool to do so. So I built a little desktop app with Tauri. It handles batch downloads without eating all your RAM (vs an Electron app) and lets you resume if something goes wrong.

\n
\n\"Setup\n
Setup instructions for Cloudinary Backup tool requesting cloud name, API key, and API secret credentials to download photos.
\n
\n
\n\"The\n
The Cloudinary Backup desktop app browsing assets before export
\n
\n
\n\"Batch\n
Batch download in progress in the Cloudinary Backup app
\n
","metadata":{"title":"Cloudinary Backup Tool","date":"2025-01-15T05:00:00.000Z","github":"https://github.com/ejfox/cloudinary-backup-tool","tech":["Tauri","Rust","TypeScript","Vite"],"modified":"2025-08-13T15:58:10.000Z","tags":["tools","photography"],"words":87,"images":3,"imageDetails":{"total":3,"cloudinary":3,"withDimensions":0},"links":3,"codeBlocks":0,"headers":{},"toc":[],"type":"post"}},{"slug":"projects/creative-interfaces","title":"Hand-Tracking & Generative Art","html":"

These are my attempts to make computers that respond to the way humans actually move and gesture—exploring the design space between human intuition and machine capability. Each project pushes beyond traditional HCI constraints to find new interaction models. Some of these work better than others, but they all represent steps toward interfaces that feel more like extensions of thought than barriers to it.

\n
\n\"Playing\n
Playing Super Mario Bros. through the hand-tracking MIDI controller
\n
\n
\n\"Turning\n
Turning a backyard crow's movements into music
\n
\n

Projects

\n
    \n
  1. hand-midi-controller: Professional hand tracking MIDI controller with TouchDesigner integration. Converts hand movements to MIDI with detailed performance timing breakdowns and memory analysis for creative interface exploration.
    2. \n
  2. handtrack-websockets: Real-time gesture streaming pushing the boundaries of browser-based interaction.
    3. \n
  3. flipper-generative-art: Real-time generative art and animated patterns for Flipper Zero with 10 gradient types, Floyd-Steinberg dithering, and smooth 30 FPS animations. Finding art within the constraints of 64KB RAM.
    4. \n
  4. ps5-tmux is a little tool to make it easier to vibe-code; just move between panes with your PS5 controller and accept changes with the X button, or hit the middle button to enter dictation mode.
    5. \n
","metadata":{"title":"Hand-Tracking & Generative Art","date":"2024-07-01T04:00:00.000Z","modified":"2025-08-13T15:46:54.000Z","tech":["Computer Vision","MIDI","WebGL","Hand Tracking","Generative Art","Interactive Media"],"state":"deployed","aiInvolvement":"ai-assisted","tags":["creativity","interfaces"],"words":186,"images":2,"imageDetails":{"total":2,"cloudinary":2,"withDimensions":0},"links":6,"codeBlocks":0,"headers":{"h2":1},"toc":[{"text":"Projects","slug":"projects","level":"h2","children":[]}],"type":"post"}},{"slug":"projects/photography-tools","title":"EXIF Printer","html":"

I love being able to reference the technical details when editing photography—aperture, shutter speed, ISO, all the little choices that go into making a photograph. But most photo prints completely ignore this data, treating it like metadata to be hidden away in EXIF tags that nobody ever sees. So I made a little app to add the EXIF data to the margins of prints you might get from CVS, Walgreens, or Wal-Mart (don't use them for photo prints, I learned the hard way.)

\n
\n\"EXIF\n
EXIF Photo Printer application displaying five family photos with options to regenerate layouts or save prints at 4x6 size.
\n
\n
\n\"The\n
The EXIF Photo Printer marketing site
\n
\n

exif-printer: Creates printable photos with camera settings displayed in the borders - a photographer's tool that treats metadata as part of the art.

","metadata":{"title":"EXIF Printer","date":"2024-04-01T04:00:00.000Z","modified":"2025-08-13T15:20:07.000Z","tech":["Photography","EXIF","Print Design","Web Design","Camera Metadata"],"state":"deployed","aiInvolvement":"ai-assisted","tags":["photography","tools"],"words":132,"images":2,"imageDetails":{"total":2,"cloudinary":2,"withDimensions":0},"links":3,"codeBlocks":0,"headers":{},"toc":[],"type":"post"}},{"slug":"projects/coach-artie","title":"Coach Artie","html":"
\n\"Coach\n
Coach Artie landing page — \"Why Coach Artie?\"
\n
\n
\n\"Coach\n
Coach Artie interface
\n
\n
\n\"Coach\n
Coach Artie in conversation
\n
\n
\n\"Coach\n
Coach Artie ecosystem view
\n
\n

Coach Artie v1: Discord Studio Assistant

\n

I created Coach Artie to serve as our Discord studio assistant. He's evolved into a powerful AI tool with persistent memory and tool usage - helping our team coordinate and manage creative workflows.

\n

Coach Artie v2: Production AI with MCP Tools

\n

Currently in development.

\n

Memory & Consciousness Analysis

\n

Explores AI consciousness emergence through memory patterns with scientific rigor applied to existential questions. This research component investigates how persistent memory systems in AI assistants might lead to emergent conscious-like behaviors.

\n

Related Content

\n","metadata":{"title":"Coach Artie","date":"2023-03-21T04:00:00.000Z","modified":"2025-08-26T19:52:50.000Z","url":"https://github.com/room302studio/coachartie","tech":["Node.js","Discord.js","OpenAI API","PostgreSQL","MCP","AI Consciousness Research"],"state":"evolved","aiInvolvement":"ai-enhanced","context":"collaborative","tags":["programming","ai","product","nodejs"],"words":143,"images":4,"imageDetails":{"total":4,"cloudinary":4,"withDimensions":0},"links":7,"codeBlocks":0,"headers":{"h2":3},"toc":[{"text":"Coach Artie v1: Discord Studio Assistant","slug":"coach-artie-v1-discord-studio-assistant","level":"h2","children":[]},{"text":"Coach Artie v2: Production AI with MCP Tools","slug":"coach-artie-v2-production-ai-with-mcp-tools","level":"h2","children":[]},{"text":"Memory & Consciousness Analysis","slug":"memory--consciousness-analysis","level":"h2","children":[{"text":"Related Content","slug":"related-content","level":"h3","children":[]}]}],"type":"post"}},{"slug":"projects/ccrb-clusters","title":"Clusters of NYPD Misconduct","html":"
\n\"Force-directed\n
Force-directed network graph of NYPD civilian complaint clusters showing officers connected by shared misconduct patterns
\n
\n
\n\"The\n
The published Gothamist/WNYC investigation built on this network analysis
\n
\n

In 2021, I got my hands on a massive NYPD complaint dataset from Gothamist/WNYC and immediately went down a data rabbit hole. Built a network of 29,915 officer nodes with 159,671 relationships using Neo4j and Gephi (love-hate relationship with Gephi, but unrivaled for networks).

\n

When Gothamist reporter George Joseph interviewed victims, they confirmed what the network showed - certain officers were \"catalysts\" who other officers \"followed\" into misconduct. Math and reality matched up! I documented everything in excruciating detail so others could replicate it for their own police departments.

","metadata":{"title":"Clusters of NYPD Misconduct","date":"2021-06-15T04:00:00.000Z","modified":"2025-08-26T19:52:56.000Z","url":"https://gothamist.com/news/mapping-clusters-nypd-officers-repeatedly-accused-misconduct","tech":["Neo4j","SQLite","Gephi","Network Analysis","Cypher"],"featured":true,"state":"deployed","aiInvolvement":"human-only","context":"client","tags":["data","nypd","journalism","database"],"words":114,"images":2,"imageDetails":{"total":2,"cloudinary":2,"withDimensions":0},"links":2,"codeBlocks":0,"headers":{},"toc":[],"type":"post"}},{"slug":"projects/covidcast","title":"COVIDcast","html":"

During the early COVID-19 pandemic, I worked with Carnegie Mellon University's COVIDcast project, prototyping data visualizations and user interactions to help the public understand hospitalization, transmission, and movement trends across the country.

\n
\n\"COVIDcast\n
COVIDcast dashboard: weekly case, hospitalization, and death rates above an interactive county-level map
\n
\n
\n\"Population\n
Population density map of the United States with circles sized by population, colored white to red indicating concentration in the Midwest and Northeast.
\n
","metadata":{"title":"COVIDcast","date":"2020-04-24T04:00:00.000Z","modified":"2025-08-26T19:53:05.000Z","url":"https://campustechnology.com/articles/2020/04/24/carnegie-mellon-maps-offer-more-data-for-covid-19-forecasting.aspx","tech":["Data Visualization","JavaScript","COVID-19 Data","Public Health"],"state":"deployed","aiInvolvement":"human-only","context":"client","tags":["covid","dataviz","javascript","health"],"words":69,"images":2,"imageDetails":{"total":2,"cloudinary":2,"withDimensions":0},"links":3,"codeBlocks":0,"headers":{},"toc":[],"type":"post"}},{"slug":"projects/nbcnews-specials","title":"NBC News Specials","html":"

During my time at NBC News, I worked on the Foundry team, creating bespoke longform articles in collaboration with journalists across the organization. These specials focused on unique approaches to video, data visualiation, and storytelling on the web.

\n

Does America Need Another Prison? (2018)

\n
\n\"A\n
A map comparing the location of prisons and coal mines in the southeastern United States
\n
\n

Post Bail (2017)

\n

The Making of an Astronaut (2017)

\n

Journey of a Bullet (2017)

\n
\n\"The\n
The introduction to the Journey of a Bullet story
\n
\n

Donald Trump's First 100 Days (2017)

\n

How Trump Won (2016)

\n
\n\"A\n
A table a map showing the counties Donald Trump swung most
\n
","metadata":{"title":"NBC News Specials","date":"2016-12-14T20:01:50.000Z","modified":"2025-08-13T16:01:12.000Z","url":"https://nbcnews.com/specials","context":"client","tags":["dataviz","journalism","webdesign","elections"],"words":104,"images":3,"imageDetails":{"total":3,"cloudinary":3,"withDimensions":0},"links":10,"codeBlocks":0,"headers":{},"toc":[],"type":"post"}},{"slug":"projects/nbc-big-board","title":"NBC News Big Board","html":"

I spearheaded an internal effort to develop new D3-powered mapping software for visualizing election maps and results live - the touchscreen app wielded by Steve Kornacki known as the \"Big Board\". Built for real-time county-level election analysis.

\n
\n\"The\n
The Big Board UI: 2016 Pennsylvania presidential results beside an interactive county map
\n
\n

2016 NBC News County-Level Election Maps

\n
\n\"Planning\n
Planning board — printed election-graphic studies and notes pinned to the wall
\n
\n
\n\"Broadcast\n
Broadcast setup — the Big Board on a control-room monitor beside a waveform scope
\n
\n
\n\"Printed\n
Printed design studies for the on-air election graphics, spread across a desk
\n
","metadata":{"title":"NBC News Big Board","date":"2016-11-08T05:00:00.000Z","modified":"2025-08-13T16:24:29.000Z","tech":["D3.js","JavaScript","Touch Interface","Election Data"],"featured":true,"state":"deployed","aiInvolvement":"human-only","context":"client","tags":["d3","dataviz","javascript","elections"],"words":95,"images":4,"imageDetails":{"total":4,"cloudinary":4,"withDimensions":0},"links":6,"codeBlocks":0,"headers":{},"toc":[],"type":"post"}},{"slug":"projects/dataproofer","title":"Dataproofer","html":"

Along with Gerald Rich and the Vocativ data team, I received a Knight Prototype Fund grant to create Dataproofer - an open source tool to check data for reliability, missing data, and outliers. Built as an Electron desktop app for journalists.

\n
\n\"Dataproofer\n
Dataproofer running its suite of data-quality tests against a CSV
\n
\n
❯ dataproofer Criterion\\ Channel\\ videos\\ v1.2.0-20190621\\ -\\ v1.2.0-20190621.csv\n \ntotal rows 5035\nrows sampled 1259\n \nMissing or duplicate column headers: passed\nNumeric cells: info\nString cells: info\nEmpty Cells: warn\nSpecial Letters & Characters: warn\nDuplicate Rows: passed\nPotentially missing rows: passed\nWords at their character limit: failed\nInteger at its SQL upper limit: passed\nSummed integer at its upper limit: passed\nSmall integer at its SQL upper limit: passed\nBig integer at its SQL upper limit: passed\nOutliers from the mean: info\nOutliers from the median: info\nInvalid coordinates: passed\nVoid coordinates: passed\n \n75%\n9 tests passed out of 12\n \n### PROOFED ###
","metadata":{"title":"Dataproofer","date":"2015-11-03T05:00:00.000Z","modified":"2025-08-26T19:53:33.000Z","url":"https://github.com/dataproofer/Dataproofer","tech":["JavaScript","Electron","Data Quality","CSV"],"featured":false,"state":"deployed","aiInvolvement":"human-only","context":"collaborative","tags":["data","javascript","tool","journalism"],"words":157,"images":1,"imageDetails":{"total":1,"cloudinary":1,"withDimensions":0},"links":3,"codeBlocks":1,"headers":{},"toc":[],"type":"post"}}],"_links":{"self":"https://ejfox.com/projects.json","html":"https://ejfox.com/projects"}}