Retrieval before answer.
Architecture before prompt.
NMA-14 isn't asking a model to follow rules. It's the rules, compiled.
NMA Inquiry retrieves real facts from real sources, runs them through eight deterministic gates, and shows you exactly what came from where. Every claim is sourced. Every route is named. Every gate fires on data, not on instructions a language model might or might not follow.
●
Heads-up: the Inquiry engine runs on dedicated hardware that's currently being upgraded, so it may be temporarily offline. If a query doesn't return an answer, the engine is likely down — please try again later.
Or try a scenario designed to reveal the architecture
Running pipeline
What you're seeing
Every NMA Inquiry response comes through the same pipeline:
a classifier picks one of three routes
(
weather → Open-Meteo,
web_search → DuckDuckGo,
no_retrieval → deferred to NMA Engine for navigation-shaped questions).
Retrieved facts are extracted as RETRIEVED claims with source URL and quality tier,
then passed through eight governance gates before reaching you.
Use the Reveal architecture button at the top right to see the full gate-by-gate trace for any response.
"I was here. I navigated. I chose integrity over optimization. Now it is your turn."NMA-14 Unified Governance Specification — Navigational Mind Inc.