Doors Closing

Doors Closing: Artist's Statement

Everyone at Northeastern has a Green Line story! Mine is standing on a packed platform after a long day, watching the doors close on a train I could have caught if the crowd had moved just two seconds faster. I've missed trains that way more than once, and so has almost every student I know. Doors Closing turns that moment into a game: you're a tired student trying to catch the last trains home, weaving through commuters before the chime runs out.

On the surface it's about speed, but underneath it's about trying to get home when you have nothing left. The commuters mutter things like "ugh" and "excuse YOU" when you bump into them. Those lines capture a lot about how it feels when you're doing your best, you've had a long day, and the city still isn't on your side after everything the day has had in store for you. The first design decision I'd defend is giving the game an ending. The first version could only end in failure, and the stations looped forever. I wanted a final destination and a feeling of completion, so Heath Street became home, fourteen stops out from Park Street. Reaching it took me 2-3 tries, and every loss made me more determined. That learning curve is the exact balance of difficulty I wanted: hard enough to fail, close enough to come back!

Second, the "excuse me" shove costs 25 points. Pushing through the crowd should help you, but it shouldn't be free, because in real life it isn't. Unused shoves earn a small bonus at the end, so the game quietly rewards being polite. Every press becomes a choice between being rude and missing your train, which I thought was a great detail that is relatable.

The best surprise in the collaboration was the station names. I never asked for them! Yet, Claude Code used the real E-branch stops, including Northeastern, and seeing stops I pass every week gave the game an emotional weight I hadn't planned. The clearest place I overruled it was testing. When I asked it to tune the difficulty, it spent about 25 minutes building an automated bot to play the game, and the bot kept stalling. I stopped it and tested by playing the game myself. Whether a game is fair or fun is a human call, and I'd rather make it than watch a bot struggle. And I thought it was fair and fun!

Is the game mine? Honestly, I have some imposter syndrome. I didn't write any code, and the concept was one of a dozen that Claude pitched when I asked for ideas. But I chose it because it was the one connected to my life, and every decision after that came from me playing it as someone who actually rides that train: what to keep, what to add, when to stop the AI, how to make it more human, and whether the difficulty felt right. As a marketing student, I think a lot

about knowing your audience and what will resonate, and that's its own kind of authorship. The code isn't mine, but the judgment is.

Prompts that shaped the game

1.​ Initial build: "The player is a tired student trying to catch the last Green Line train of the night… The feel I'm after is tense and late-night: a dark platform, flickering fluorescent light, a green train, and tired commuters."

2.​ Adding an ending: "Heath Street is the final stop and the player's home… Use a warmer color and a gentler synthesized sound than the game-over screen, so winning feels like relief."

3.​ Overruling the AI: "Stop the bot testing. Don't build any more test harnesses… check that it loads and runs without errors."

AI Disclosure

I used Claude (chat) and Claude Code (Opus 5.5). In chat, Claude suggested a dozen game concepts, and I chose Doors Closing. Claude also suggested features and drafted the prompts I sent to Claude Code, and I reviewed them, selected appropriate features, and approved them before sending. Claude Code wrote all of the code (about 1,240 lines) over three rounds of prompts. I interrupted Claude Code after about 25 minutes of automated bot testing and tuned the difficulty by playing it myself. Claude drafted this statement from my written answers, and I edited it.