Eleven systems, still running.
Every one of these started with a problem somebody was solving by hand. All of them are custom-built, in use today, and still maintained by us. A portfolio should be work that survived contact with real users, and this did. It runs from a media engine holding tens of thousands of files, through a living family archive with searchable interview transcripts, down to a booking page for one room in Greenville, and we'd take on any of those. Better still, almost all of it is live and free to try, so don't take our word for anything: click through and use it.
Shutterfield
A full media library engine for organizations that outgrew the shared drive.
Most teams outgrow the shared drive long before they admit it. Files pile up under names like IMG_4872.jpg, nobody remembers which event they came from, and finding one photo means asking the person who took it.
Shutterfield replaces that with a real library. Upload photos and video in bulk, then filter by orientation, media type, location, time of day, even the weather when the shot was captured. Rename on import, tag everything, and share privately, publicly, or through a branded portal on your own domain. Trim and arrange video, publish straight to YouTube, and manage every asset from one place.
It's the best example of what custom software means to us: a system shaped around how one operation works, then hardened until it can run unattended. Nobody started from a template or dropped your logo on a starter kit.
It's also built for a crew of one. If you fly a drone for a living, or want to, point the desktop app at a folder and every card you dump imports itself: duplicates caught before a byte uploads, masters kept safe on storage priced like storage. DJI flight telemetry rides along automatically, full flight paths with altitude and a flag for anything over 400 feet, plus the weather at the moment of capture on every clip. And the same engine can wear your brand on your own domain and take your bookings, so the library, the website, and the calendar are one system instead of three subscriptions. Big Sky 30, next on this page, is exactly that arrangement.
It's also the best catalog we have of what we can do, which is why the rest of this section reads as the problems it solves rather than a feature list.
Nine problems this solves
All of it running in production today.
- Nobody can find the thing. Shutterfield filters a library across more than thirty dimensions at once, and they compose freely: type, place, tags, a year, a month, one calendar day, every August, the hour of the day, dawn versus sunset, orientation, altitude, duration, which import it arrived in, who owns it, and free text on top. Any combination saves as a view you click once, so the folder tree somebody set up in 2019 stops being the only way in.
- The data you need is already out there, and nobody's joining it. Every geotagged item in Shutterfield gets reverse geocoded to a city, state, and zip, then enriched with the historical weather at the exact moment it was captured, so "every sunny shot above ninety degrees" is two clicks. DJI drone telemetry gets parsed out of subtitle files into full flight paths with altitude and distance. Nobody typed any of that in; it was joined once and left to do its work.
- Getting it in is the hard part, and it's where most projects die. Shutterfield imports local folders with MD5 duplicate detection before a byte uploads, Google Takeout with all its sidecar metadata intact, and entire YouTube channels. Multi-gigabyte files upload in chunks that survive a dropped connection, HEIC converts on the fly, and RAW files stay linked to their JPG siblings. Migration off whatever you're leaving is real engineering, it decides whether anyone adopts the new system, and it's part of the build with no surprise invoice for it.
- The work leaves the system to get finished. Video gets edited inside Shutterfield on a real multi-track timeline: split, trim, layer unlimited audio over the original, scrub waveforms, autosave the draft so a closed tab doesn't cost the session. Export renders in the browser through FFmpeg compiled to WebAssembly, or hands off to a cloud queue that notifies you when it lands. Wherever your workflow currently exports to a second tool and re-imports the result, that round trip is a place where versions drift and hours disappear.
- Publishing is somebody's whole afternoon. Connect a YouTube channel once, then queue uploads straight from the library, pick playlists, watch status, and filter for what has not been published yet. Public albums, tokenized share links, posts, and one-tap cross-publishing to another network all come off the same records. Post it once and it's live everywhere it should be.
- You're renting your own audience back from somebody. Shutterfield serves white-label portals: branded gallery sites on other organizations' own domains, powered by the same library underneath. It auto-generates statically cached city landing pages with sitemaps for the public content, watermarks shared and downloaded media through a per-user visual editor, and gates full-resolution downloads with a complete audit trail. It all runs under your name on your domain, with your traffic staying yours and nothing at the bottom of the page pointing at a competitor.
- Who's allowed to see this, and can you prove it. Roles and capability flags, teams with invited members, co-ownership of records across accounts, and row-level security enforced in the database itself, down where a UI checkbox can't reach. Downloads are permission-aware and every one is logged. This stops being a nice-to-have the moment an auditor or an insurer asks a question, and it's very hard to retrofit.
- Built for the size you'll grow into. The gallery virtual-scrolls across tens of thousands of items without stuttering, and album counts come from a trigger-maintained cache table, so nothing gets recounted on every page load. Plenty of software feels fine on demo data and falls over by year three; this was built for year ten from the start.
- The browser isn't always the right place for it. Shutterfield also ships as a Windows and Mac desktop app: point it at a camera card or a project folder and new files import themselves, uploading in four parallel parts to use the whole uplink, with a local SQLite library, native EXIF and HEIC handling, and native FFmpeg for fast exports. It even rotates your wallpaper from an album on a schedule, from the system tray, with every window closed. When a job belongs on the machine itself, we build that piece too.
Big Sky 30
A fully branded public site that runs itself. Everything you see is powered by Shutterfield: upload to the library, add it to a public album, and it is live.
Big Sky 30 documents the everyday of this part of North Texas: drone footage over the lake, local events, the ordinary afternoons that turn into the good ole days once enough time passes. It's a public archive, free to browse, and it grows every week.
It's worth showing here because of how little running it takes. Nothing custom sits behind Big Sky 30. It's Shutterfield on its own domain, wearing its own brand. To publish, you upload footage to the library and add it to a public album. Nothing gets exported, re-uploaded, or copied into a second file that drifts out of sync.
The rest of the site is yours to run without a developer in the loop. You edit page copy inline and it goes live the moment you save. The same branded site can take bookings and collect revenue, so an archive built to look back can also book the next shoot and get paid for it. That's the arrangement we build for clients: one engine holds the truth, a branded site on your own domain shows the world what you want it to see, and you never touch anything but the content.
Life Story EcoSystems
The other ninety-nine hours of a documentary biography, indexed and searchable, held for the people who are not born yet.
A documentary biography collects fifty to a hundred hours of interviews, photographs, and records. About an hour reaches the screen. The rest sits on hard drives, undated and unsearchable, until nobody left knows what is on them. Genealogy sites give you a tree. Cloud storage keeps the bytes. Neither one holds a life.
Life Story is where the rest lives. Before a frame is shot, the family gets a private link, no account required, and starts dropping photographs, letters, deeds, and cassette transfers into their own archive. Files are fingerprinted in the browser before they move, land in a private bucket that belongs to that archive alone, and come back with thumbnails, viewing copies, extracted audio, and diarized transcripts stamped to the millisecond.
Search is one box across speech, writing, scans, and names. Type a phrase and you land on the second it was said, not a file to scrub through. The creative structure and the database structure are the same object: themes, people, places, and periods linked the way a filmmaker already thinks about a life.
It is a working prototype, and the site says so plainly: seven things run in production today, the rest named as in-build or planned rather than implied. That honesty is part of why it belongs here. Built as a Data Day Studio prototype for a documentary partner, it is also the cleanest example we have of an archive that keeps the meaning, not just the files.
Six problems this solves
The files were already safe. The meaning was not.
- Most of a life never makes the cut. Fifty to a hundred hours of interviews and records shrink to about an hour on screen. Life Story holds the rest as a living archive instead of a drawer of drives nobody can search.
- Storage solved the wrong half. Backups keep the bytes; they do not keep the context. A photo of a brick house in a folder called 2015 uploads is a house until somebody can attach who mortgaged it, and why.
- Asking an eighty-year-old to create a login is where this fails. A private link is the credential. It writes into one archive only, accepts anything at any size, and refuses nothing that might later turn out to matter.
- Finding a moment meant scrubbing footage. Interviews come back diarized with a timecode on every sentence. Search returns the second a phrase was said, across speech, notes, and scanned documents, in one box.
- Genealogy, memoir, and cloud each answer one dimension. A life is lineage and narrative and media and time at once. The archive holds all four and keeps them pointed at each other.
- A company forgets faster than a family does. The same engine can steward founder interviews and why a decision was really made, so institutional memory survives retirements instead of walking out the door.
Reserve My Spot
Members hold their place from home, so nobody drives over to sit in a lobby.
Membership spas sell access to a handful of stations. Red light panels, a wave massage bed, tanning beds, a sauna. The popular ones always draw a line, and the only way into that line is to drive over and wait. Members burn an afternoon guessing, and the front desk fields the same phone call all day.
Reserve My Spot puts that line on the member's phone. They pick a service, see what the wait looks like right now, and request a spot for when they plan to arrive. The desk watches the request land and accepts it, and the member gets a position and a time. The wait estimate comes from the spa's own session lengths and station counts, simulated forward through the current queue, so it stays honest as the afternoon fills up.
The part off-the-shelf booking software can't do is tell staff who's on the way. With opt-in location sharing, the app checks a member's live driving ETA against the minute their station frees up. Once the grace period the business sets runs out, the desk gets one clear choice, bump them back or release the spot, in place of an awkward wait beside an idle station. Three surfaces run off one live database, so a request appears at the desk the instant it's made, and a call-up reaches the member's phone before they get to the counter.
Sell On Hand
Live stock on a phone, paid at the booth, with no hardware between the seller and the sale.
Direct sellers, chefs, and table vendors already know what they have on hand. The hard part is keeping that count honest while people are buying, and taking card or cash without a separate till. Sell On Hand is the toolkit for that: inventory, a shopper-facing storefront, card payments, cash confirmation codes, receipts, reporting, and customer texts, all on the phones already in the room.
A shopper opens the vendor's link, sees what is actually left, and pays on their own device. Stock drops the moment the sale clears, so the next person never sees a ghost item. The vendor can watch live carts in realtime, push a discount to a basket that needs a nudge, and run cash when the card reader is the wrong tool.
The first customers were Jordan Essentials consultants. Their branded host is JE with Me at jewithme.com: the same product, wearing the JE skin, with a consultant directory and storefronts scoped to that brand. Sell On Hand is the platform anyone can join; JE with Me is the proof that one engine can wear another company's name without becoming a second build.
Church Street Studio
Direct-to-consumer booking that skips the platform fees.
A creative space in Greenville, Texas, rented out for remote work, events, and creative sessions starting at $41.
The booking runs through the studio's own site, so the fee that would have gone to a marketplace stays with the business. It also fills a gap the big platforms leave: micro-rentals by the hour or the afternoon, the kind of short booking Airbnb was never built to list.
The availability calendar syncs automatically with Airbnb, so a listing can go out on both channels without double-booking or hand-updating.
This is the smallest kind of thing we build, and it's worth showing next to the big ones, because the machinery underneath is the same. A calendar, a payment, and a confirmation email with the right name on it is most of what a clinic or a salon needs too.
69.studio
A podcast platform where the show owns the audience, plus a complete single-show listener app built on the same machinery.
69.studio plays any show from one link. Paste a YouTube channel or a podcast feed and the page builds itself: artwork, the full episode list, a player. An episode that exists as a video and as audio is one episode with a Watch or Listen toggle, and your spot carries across devices to the exact second. Search runs on the words themselves: transcripts are indexed, so a typed phrase returns the episode and the second it was said.
We've also built the deeper version: a dedicated app for a single show, with the sponsors on the inside. Sponsors live in rooms of an illustrated house, and every tap on that house is counted, filtered for bots, and attributed. Each sponsor signs into a dashboard of their own: verified visits and calls, a map of where they came from, and the performance of the social posts that mention them, day by day.
Premium works without asking anyone to cancel anything. A Patreon subscriber links their account and membership verifies automatically; a Substack subscriber verifies by email. The paywall is enforced in the database itself, not just hidden in the page, so a locked episode advertises that it exists while the media stays locked.
Together they are the answer to a question every show eventually asks: why does the audience we built belong to the platforms we post on?
Legal to Fly
FAA Part 107 drone exam prep that grades you the way the FAA does.
Anyone flying a drone for money needs the FAA Part 107 certificate, and the prep market sells video courses for $149 to $249. Legal to Fly is the precision alternative: a question bank where every question carries the FAA's own ACS element code and the regulation citation it came from, with three answer choices, because that is what the real exam gives you, not four.
Practice sessions score you by subject area. A daily drill uses spaced repetition, so the questions you miss come back until you stop missing them. The timed mock exam runs 60 questions in two hours at the real 70 percent pass mark, and when you miss, it hands you the list of ACS codes you missed in the same format the FAA prints on your actual test report. Sessions even sample questions in the same proportions the FAA weights the exam.
It was built the way we build for clients: the answers are verified against the fetched text of the regulations themselves, not against somebody's course notes, so when the FAA amends a rule, every affected question can be found and fixed by its citation.
Three small products, running on their own.
The next three are consumer products in their own right: a family stats logbook, a daily word game, a playable music archive. They also happen to be the clearest small demonstrations of the thing we sell: record something once, and every number derives itself. All of them are free. Click in and play.
42 Stats
A private logbook for the domino game, where one entered match updates every number in the house.
Texas 42 is the state's official domino game, and families play it the way families do: often, competitively, and with standing arguments about who is actually better. 42 Stats settles them. Recording a match takes a date, four names, and a few 7-point scores. Everything else derives itself: standings, head-to-head records, partner-pair ratings, average margins, form, a win percentage timeline, and BidWAR, a net-bidding metric made for the game.
Nothing derived is ever stored. Every stat is computed fresh from the raw matches at the moment you look, so one saved match instantly changes every page. And the permission model runs through the math itself: a viewer's standings are computed only over the matches they are allowed to see, enforced in the database rather than in the page.
It is the cleanest small example of the thing we build most often for businesses: record the event once, and the reporting stops being anyone's job. Browse the demo league without an account and see for yourself.
Suffering Jukebox
A community catalogue you can browse, play, and put on air in a room.
It started as one artist's discography charted by YouTube views. It is now a public catalogue anyone can grow: import a whole artist from YouTube, drop in a single song, or bring a Spotify library across, then choose whether it lands in the public Explore Artists grid or stays private in My Jukebox.
Browse by artist or by playlist. Open an artist and the albums line up by year with real view counts on every release and every track. Hit play and the queue runs with time-synced lyrics beside the video. Build playlists, shuffle a whole catalogue, and share lists that open without an account.
The Online Jukebox is the room version. A host puts a station on air; anyone who opens the link joins with no sign-in, watches the same song at the same second, browses the catalogue, and adds to a shared queue everybody sees. When the room goes offline, the last running order is still there to send into your own player.
On a phone, YouTube stops the moment the screen locks. Background play is the answer we built for that: artist-approved or personal audio keeps going on the lock screen, over CarPlay, and over Android Auto, while discovery and normal listening still use YouTube. Sign in and the charts can redraw against your own plays instead of the world's.
SixGuess
A family-friendly Wordle competition, automated into a website.
It started with three family members texting each other their daily Wordle results in late 2023. That turned into a Discord server with a custom bot: players uploaded screenshots, the bot read them with OCR, and the results fed a Google spreadsheet that fed a dashboard, refreshing nightly.
It worked. It was also a lot of steps. You had to install Discord, find the right channel, and remember to upload a screenshot every single day. Seventeen players logged more than 3,700 games that way before it ran its course.
SixGuess collapses that whole pipeline into one place. Open the app, play the puzzle, and your result is on the leaderboard. It's the same competition and the same family, with the friction gone.
It's on this page for a reason that has nothing to do with word games. Somebody was running OCR on screenshots and reconciling a spreadsheet by hand every night. That's exactly the kind of manual routine we get hired to erase. Even the screenshot pipeline survived the upgrade: paste a screenshot of a Wordle you played elsewhere and the app reads the tiles into a real result.
The stats run six tabs deep: rankings sortable seven ways, guess distributions, head-to-head comparisons across any set of players, a browser for every guess anyone has ever made with wildcard word search, and analysis of which opening words actually earn their reputation. A win means you beat everyone that day, not just the puzzle, and the streaks are scored accordingly.
Yours could be next.
A prototype is $500: a working version you can try. If you want to take it to production, we quote that build next.