CHOOSESecurity & safety
Shadow Gate Model Analysis
This project analyzes past posts using a TypeSafe Jev model to categorize assertions into cause, prediction, or advice.
Build log 1/7 — shadow gate. We ran 326 of our own past X posts through a decision-only model (TypeSafe Jev) with one question: does the text assert a cause, a prediction, or advice? 316 none, 6 causal, 3 predictive, 1 advice. p50 232 ms, 174k input tokens. Feed unchanged so far.
