What we claim — and what we don't

• Deckle™ — validated limit It does not improve already-polished professional writing.
• Deckle and Strop™ — validated limit Long documents are cut at 24,000 characters, and the record stamps the source as truncated. What you did not find in a truncated passage is unchecked, not absent.
• Geiger™ — validated limit A rare structural pattern — text that reads clean but forecloses by design — is flagged as a caution, not a verdict. Geiger has not confirmed this pattern against enough real cases to score it, and it says so rather than pretend certainty.
• Geiger — validated limit Geiger reports a band and the observed range across repeated readings, never a single confident number. On a close call, it tells you the call is close rather than manufacture a decisive-looking score.
• Strop — a refusal Strop does not write your policy. The connections between claims are the policy, and a record certifying work the instrument did would be worse than no record.
• The narrower claim Nothing here is ungameable. Gaming language becomes far more difficult against a measurement the writer does not control.
• Attribution The record shows what came from you, what was proposed, what you chose to take forward, and why.
• Pilot posture Geiger is spec-grounded on some questions, not yet specimen-validated on all. Where it does something today, it is live. Where it doesn't, it is a proposal you are being asked to shape. A pilot is not a proof.