FDAG is the memory-over-time layer behind the public trail.
Most AI memory systems retrieve something that looks relevant and place it back into context. That can be useful, but it leaves a harder question unanswered: did that memory earn influence?
FDAG treats memory as an evidence trail, not a scrapbook.
The point is not to stuff more history into prompts. The point is to preserve what happened across runs, separate witnessed events from guesses, and keep influence gated until evidence survives controls.
In the Probably AGI stack, FoxEIS tests inference events. FDAG tracks evidence over time. PX maps code and repo context. YouDAG records episodic action trails. Together they form a public-safe silhouette of a larger substrate for human-AI symbiosis infrastructure.
The boundary matters.
FDAG does not claim AGI, production steering, hidden model memory, or automatic authority. A memory can exist without earning influence. A pattern can look promising without becoming a rule. A negative result can still be useful evidence.
That is the point: memory should be earned, not assumed.
This public page does not include exact coordinates, prompt packs, kernel details, source code, private logs, or reproduction paths.