Build Log
Aug 2026 build
Shipped /de-slop, and proved the wedge instead of claiming it
The judgment layer above the em-dash remover, split into three tiers by determinism rather than topic, then tested head to head against a blank model on the same draft.
12/12 planted tells caught, 0% of a clean draft changed
Claude Code
The blank model caught 8 of 12 planted AI tells, fabricated a customer quote inside a blockquote, and rewrote 29% of an already-clean draft while reporting zero findings. The skill hit 12 of 12, left protected regions intact, and changed nothing. The honest limit went into the lesson itself: the blank model is competent. The wedge is not that a model cannot find slop. It is that it does not know what it may not touch, and it will not stop.
Related project: Zero to Automated