Quality at Volume

The temptation at volume is to relax quality standards — "we need 50 assets by Friday, so the critic gets one round instead of three." This is a false economy. A batch of 50 low-quality assets creates more rework than a batch of 40 high-quality assets with 10 in the revision queue.

The batch quality protocol: every asset in the batch passes through the same critique pipeline as a single asset. No exceptions. No "quick pass" mode. The efficiency gain comes from parallelization and templates, not from lowering the quality bar. If the batch is too large for the timeline at full quality, reduce the batch size — not the quality threshold.

Batch-specific quality patterns: the critic should evaluate both individual quality AND batch coherence. After every 10 assets in a batch, run a cohort check — do the last 10 look like they belong together? Batch generation can drift faster than single-asset generation because the model processes many requests in rapid succession, and each micro-variation compounds with the next. The cohort check catches this before 50 assets drift off-brand.