06 / Deterministic generative music

Turn a place into the same piece of music every time.

A browser composition engine that maps normalized coordinates to a versioned musical plan, while treating listening quality as a separate gate from deterministic correctness.

Product brief

Product
Mapsonata
Status
In progress
Target user
A curious listener who wants a place to produce a distinctive, repeatable piece of music rather than an opaque one-off generation.
My role
Product definition, deterministic composition contract, browser audio implementation, fixture design, listening gates, failure analysis, and cost boundaries.

The problem

A map-to-music experience fails if the same place changes arbitrarily, nearby places collapse into one sound, or technical determinism is mistaken for a piece worth hearing.

01 / Workflow

The product, end to end.

Follow the product from first input to a reviewable outcome.

Mapsonata product flow
  1. 01Enter normalized coordinates
  2. 02Build a canonical place key
  3. 03Generate a versioned composition plan
  4. 04Render with Web Audio
  5. 05Compare geographic fixtures
  6. 06Pass a human listening gate

02 / Decisions

Architecture with a reason.

Vite · TypeScript · Web Audio · Versioned composition plans · SHA-256 fixtures

01

Plan before playback

Coordinates map to a serializable composition plan first, keeping deterministic geography separate from browser rendering.

02

Version the mapping

Canonicalization and composition rules carry explicit versions so a changed algorithm does not silently rewrite earlier outputs.

03

Listening is a separate gate

Hash stability can prove repeatability, but only a directional listening pass can decide whether the music is distinct and coherent.

Failure paths

What can go wrong is part of the product.

  • Equivalent coordinates produce different canonical plans
  • Distinct fixtures collapse into musically indistinguishable output
  • The browser cannot initialize or resume audio
  • A technically valid candidate fails the listening rubric

Verification

What supports the claim.

  • Stable same-place composition plans
  • Twenty-four geographically distinct fixture plans
  • Deterministic digest checks
  • The first engine candidate was rejected by the listening gate

Portfolio boundary

Complete enough to review. Honest enough to trust.