Musephoria

What needed solving.

A simple give-feedback-to-get-feedback exchange rewards the cheapest review that satisfies the rule, not necessarily a useful critique.

1. 01Track + timed prompts
2. 02Designed review incentives
3. 03Feedback quality hypothesis

Simplified from the implementation and project notes.

How it fits together.

The proposed mechanism tied structured questions to audio playback, used model scoring to influence reputation rather than block every submission, and weighted recipients’ ratings by standing. Arbitration was designed for contested cases.

Earlier versions explored PlanetScale, Firebase, Supabase and a Postgres/Zero design with Solid, Hono and Bun. A later July 2026 repository revision uses Next.js, React and SQLite through Drizzle, with local media in development and S3 storage support. The screenshots show that later application.

The retrospective records two working hackathon pieces built with a collaborator: an inline review-writing assistant and a multi-agent quorum validator. It separately acknowledges that the wider exchange never reached a real musician trial.

The later code organizes credits and karma as append-only ledgers. Review requests and assignments have expiry rules, and domain functions accept the database and time explicitly for testing. The review cockpit uses wavesurfer.js for listening requirements and timestamped feedback; quality scoring combines a heuristic pass with an optional model pass.

Building and checking it.

Most of the work went into mechanism design and successive foundations. The published retrospective identifies that imbalance directly: each migration solved a real technical limitation while delaying the test of the core social behavior.

The lesson is to build the smallest experiment that can disprove the central assumption. In this case, that was whether two strangers would exchange useful feedback under a simple playback constraint.