One argument a week, settled by bracket.
Bracket takes one question, like “What’s the greatest Taylor Swift song?”, and settles it with a seeded, 64-entry, NCAA-style bracket. One round a day, one vote per matchup, and an argument under every pairing.
The idea
Fandoms love to argue about rankings. Bracket gives that argument a shape: each week, one question and 64 contenders, knocked out head to head over six rounds until one is left.
Tools for making brackets already exist, but they’re widgets you embed for your own audience. This is meant to be a destination: one carefully seeded bracket a week, with predictions to play along with and a debate under every matchup. The bracket is the argument engine. The daily rounds are the reason to come back.
It starts with music, and the working launch bracket is the greatest Taylor Swift song.
What makes it different
How a week works
Sixty-four entries, six rounds, one champion. Browsing is open to anyone; signing in with Google or Apple lets you vote, predict and post, under a display name you choose.
Run with Claude
Like Kantan, Bracket is built to be run by talking to Claude. Each week Claude drafts the 64 entries, with seeds and a one-line reason for each; I approve, swap and reseed.
There are two ways to run it, on one API: an admin screen, and an MCP server that gives Claude the same controls. Setting up a bracket, scheduling its rounds, reviewing suspicious votes and moderating comments can all happen in conversation.
The code is written with Claude Code, and the whole project is planned and tracked in Kantan.
What it runs on
Where it is
The bracket engine, the API, the admin tools, the MCP server, share cards and every first-version screen are built, with automated tests. The first deploy is next.
Got an argument that needs settling?
Bracket is still in development. If there’s a question you’d want settled by 64 entries and a week of voting, tell me.