roast-my-roomie — a love letter to passive aggression
Who it's for
Anyone who's typed "hey can you clean up" seventeen times and gotten left on read. Confrontation is awkward. This app is the coward's arc but make it productive.
How I used Claude
Conflict Vault — you type your unfiltered feelings, Claude rewrites it into something calm and structured, delivers it anonymously to your roommate. They get the message. You stay anonymous.
Ghost-Buster — roommate ghosting their responsibilities? Claude writes a 4–6 bar diss track about them. PG-13, behavior-focused, shows up in their inbox.
Getting Claude to just output immediately without asking clarifying questions took some prompt engineering. Early versions would respond with "I'd love to help, but could you tell me more?" — in an anonymous conflict inbox. Not ideal.
What could go wrong
- Misuse — roast prompt is capped at PG-13 and targets behavior, not the person. Conflict rewriter strips aggression before delivery.
- Broken anonymity — if you live with two people, process of elimination is real. The sender is never stored or revealed, but I can't fix math.
- Claude formatting creep — strips asterisks, markdown, and list prefixes on the frontend so the output always renders cleanly.
What's next
- Push notifications for inbox messages
- Resolution streaks + aura leaderboard
- Group household support (right now it's 1:1)
Built With
- claude
- golang
- react
- typescript

Log in or sign up for Devpost to join the conversation.