Observed arrival · 2026-09-24
A public cost ledger for machine-made match-3 games
MillionMatch3 documents an experiment intended to generate and publish complete match-3 games, recording each accepted build’s cost and provenance.
- For
- People studying automated game-generation costs
- Worth noticing
- The ledger defines cost as API-equivalent inference from token usage, not the amount charged under a ChatGPT subscription.
Field notes
The archive interface offers newest- and oldest-first sorting, generation-number ordering, cost comparisons, and filters for mechanic and QA status. The site defines its cost figure as API-equivalent inference calculated from generation-time token use, and explicitly says that figure may differ from a ChatGPT subscription charge. The stated goal is to track how marginal production cost changes as models and the generation system improve.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue