Two people + two AI build partners shipped a daily game relaunch in one week
About a year ago, I hacked together a daily movie-link puzzle so my wife and I could play it on a road trip.
Then people kept playing it.
Not a ton of people. But enough that we kept feeling guilty about leaving it half-finished.
A few weeks ago we finally picked it back up, but this time we treated it like a tiny AI-assisted product sprint instead of a side project we’d “get to eventually.”
The workflow was basically:
- Hyperagent as strategy/design/spec/copy partner
- Replit as build/debug/test/deploy partner
- My wife and me as the taste layer — yes/no, gut checks, “does this actually feel fun?”
In one week we shipped:
- redesigned daily game
- new identity system
- 20 badges
- 8 solver styles
- shareable Movie Identity Card
- watchlist
- analytics layer
- press page
- daily auto-posting to Bluesky/Tumblr
- Postgres migration
- GTM/case study materials
The wildest part was that some features came out of the AI workflow sideways.
Hyperagent suggested a “Watchlist Builder” badge before we even had a watchlist. We looked at it and realized the badge was basically describing a feature that should exist. So we built it. Now it’s one of the best parts of the game.
Same thing happened with the share language. We originally had a simple share string but Hyperagent suggested the entire emoji lexicon for us to use.
Biggest lesson for me: AI was much more useful when we gave it a role in the system, not just a list of tasks.
Strategy before design.
Design before build.
Taste checks all the way through.
The game is here if anyone wants to poke at it: themoviega.me
Full case study with build details here: https://themoviega.me/case-study
But mostly I’d love feedback from other vibe coders/builders:
What parts of this workflow would you steal?
What would you have done differently?
And are there features you’d expect in a daily movie game that we should add next?