RedHub AI
Field Guide · No. 01
Deterministic Gate · How it works
Anatomy of
a Gate
Every RedHub tool builds a weighted score — then runs it past a gate that can only make the verdict worse. This is why an 89 still fails.
89/100
Weighted score · looks like a pass
Above the 75 threshold. On the number alone, this ships.
The controls that built the 89
01Citation presentwt 20
02Source supports the claimwt 20
03Internally consistentwt 20
04Fabrication-freewt 20
05Citation reachablewt 20
The Gate · worsen-only
Comes in
Score 89 — would clear
can only lower
Fatal flaw found
1 material claim on a fabricated citation
Rule: any material claim resting on an unreachable or fabricated citation forces the floor verdict — regardless of the score. The failure is local; a high average hides it.
Do Not Ship
despite scoring 89 / 100
Gated
The score is context. The gate is the answer. A tool that only averages would have blessed a board memo built on a source that doesn't exist — the one thing that would've embarrassed you, hidden behind a good number.