AVAILABLE · 1 project slot — booking November 2026 EST · PHILADELPHIAv0.3.0
Home/Work/DragonLily Productions
SHIPPED · 2026 // client work — video production site

DragonLily Productions.

A video production studio needed a fast splash to launch on — now it's the staging ground for a full curtain-reveal reel showcase, built and tested behind a feature flag so production never breaks while the next version ships underneath it.

Role
Design & build
Timeline
3 months, ongoing
Year
2026
Stack
Astro · Tailwind
Status
Live in production
hero shot — splash & reel showcase
DragonLily Productions — live splash, with the Velvet reel showcase behind a flag
// results
100%
test coverage enforced
36
components, atomic design
30 PRs
shipped in 3 months
2
experiences, one flag
The start
01 · context

Get something live, first.

DragonLily Productions is a video production studio run by Aaron Bryan — an Emmy and Peabody Award–winning producer, director, and editor with thirteen years across broadcast, digital, branded content, sports, and live entertainment. The first version of the site was deliberately small: a logo, a one-line mission statement — “Dedicated to creating content that explores and exemplifies the spectrum of humanity” — and a contact link. Get something real in front of clients, then build.

The approach
02 · strategy

Building the next chapter without breaking the first.

The next version, internally called Velvet, is a full curtain-reveal experience: a reel showcase with category chips, reels that play inline behind an animated curtain, a fullscreen handoff to Vimeo or YouTube, and a “Behind the Curtain” sheet carrying Aaron’s bio and credits. It’s a different site, not a redesign of the old one — so it had to be built, tested, and reviewable without ever putting the live splash at risk.

The answer was a build-time feature flag. With PUBLIC_VELVET_ENABLED off, production ships exactly the original splash — byte-identical, nothing extra in the bundle. With it on, both experiences bundle together and a small runtime toggle lets a shared preview link default to Velvet while still letting a reviewer flip back to the original for comparison. Real client sign-off on a real deploy link, with zero exposure to visitors on the live domain.

What I did
03 · the build

Engineered to stay honest under review.

  • Atomic design, all the way down — 36 components split into atoms, molecules, and organisms, each with its own Storybook story.
  • Pure state, wired separately — interaction logic lives in plain, DOM-free reducer modules; the components themselves just translate DOM events into state and paint the result back. That split is what makes 100% test coverage (lines, functions, branches, statements) enforceable at every commit, not just aspirational.
  • Two full end-to-end suites — Playwright against the legacy splash and, separately, against a Velvet build — plus a docs site, password-gated the same way Storybook is, that walks the whole Velvet experience feature by feature for anyone who needs to review it without reading code.
  • Releases that write themselves — semantic-release reads conventional-commit messages on merge and cuts the version bump and changelog automatically; nothing is hand-tagged.
The outcome
04 · where it stands

Where it stands now.

Thirty PRs and about three months in, Velvet is functionally complete behind its flag — reel playback, contact form, the full mobile-first layout — with one milestone left: the flag flip that makes it the site everyone sees. Until then, DragonLily Productions’ visitors keep getting the original splash, exactly as it’s always worked, while the next version gets built and proven in the open.

// next project
Up next · client work
iCanCook →
↗

Have a build that needs this?

That's the kind of thing I do. One slot opens in November.