Playable Models
Which model each step of the playable pipeline runs. One Default config sets the floor; each AI Mode stores only what it overrides.
Only the bindings this mode authors are stored. Everything else falls through to the Default config.
Within config compares a step to its per-kind default. vs Default compares this mode to the Default config.
Cost matrix
Catalog pricing × generation parameters. Bound models (current per-kind default) are marked.
What-if: model swap
Simulated only — nothing here changes bindings or billing.
Current (Video)
no default bound
Candidate
Playable recipe (editable)
Default (all tiers)
baseThe rung every tier inherits, and the only row tier-less callers reach.
What actually served a run
Pick a recent web playable to see, per stage, which config the API resolved — config:<name>@<tier>, fallback-row:<name> or hardcoded-ladder. This is the run's own record, not a re-derivation of the selectors above.