Contributing to the playbooks

Every verdict on this site is a dated judgment, and the point of the playbooks is that they get re-scored as the frontier moves. A stage marked Avoid in September can become Draft by December — usually because verification got cheaper, not because the model got smarter. When that happens, the playbook should say so, and the person who noticed should get the credit.

How it works today

The playbooks live as one JSON file per domain in the site’s GitHub repository (data/domains/*.json). To propose a change — a re-scored verdict, a renamed blocker, an updated tool list, or a whole new domain — open a pull request that edits the relevant file. The schema is documented in data/SCHEMA.md. A verdict change should say what dissolved the blocker; “the models got better” is almost never the real answer.

What’s coming

In-app contributions: signed-in students proposing stage-level revisions with a required rationale, instructor review, and a visible revision history per stage — the September-to-December diff is the most instructive artifact a playbook can produce. Until that ships, the pull-request path is the contribution path.