What it does
You perform chest compressions by clicking in time with a beat. Tempo is scored against the rate CPR is actually taught at, and how long you hold the click maps to compression depth — hold too long and you're going too deep, release too early and you're too shallow.
Around that core loop sits the rest of the emergency sequence: check the scene, call for help, open the airway, check for breathing, decide whether compressions are even the right response. The levels are deliberately silly — a zombie in a haunted house, an alien aboard a saucer, a dragon in a castle — because the technique underneath is identical either way and people remember things they enjoyed.
Design log
These are judgment calls, not bug fixes.
The realistic version taught the wrong lesson
The scene safety check started as a thorough multi-step sequence — accurate to how it's taught, and a chore to click through every single time.
The game's own premise is that survival odds fall roughly seven to ten percent for every minute without compressions. A version that makes you spend twenty seconds tidying up the scene before you're allowed to start is quietly teaching the opposite of its own message. The mechanic contradicted the content.
Cut it to one or two clicks and a couple of seconds. You still check, because checking is real and skipping it would be worse — but the game now communicates urgency with its pacing instead of just saying it in the text.
No mandatory tutorial
The first build opened with a required training-dummy drill before you could reach a real level. I removed it and put players directly into a themed level.
A tutorial is where people quit. The compression mechanic is one click held for a moment — it teaches itself in about four seconds of play. Front-loading instruction to protect against confusion that wasn't going to happen costs you the players who never got past it.
Two builds from one codebase
The public version is generic. A separate build with organisation-specific branding exists for sharing internally, gated behind a shared password.
I work in this industry, and putting an employer's trademarks and product names on a personal project I published without going through marketing is not a call I get to make on my own. Splitting the build meant I could share the fun version with coworkers without the public one carrying anything that isn't mine.
Rather than paying for a hosting tier with password protection, I wrote an edge function to handle the shared-password gate. Coworkers visit the URL and type a password.
Known gaps
- The breath quality readout reports zero percent in the debrief. Known, unfixed.
- Possible dropped compressions on one level's artwork. I thought I'd found the cause, tested it, and had to retract that — so it's currently an open question rather than a diagnosed bug.
- The three original campaign levels have never had a confirmed clean end-to-end playthrough of the full CPR loop.
- A player-feedback pass is outstanding: the post-level quiz marks one answer wrong without ever giving the player the detail needed to answer it correctly.