Due diligence conducted by agents, against structured tools
This is Jonathan Hare's account, with its provenance shown. The person or organization named has not yet claimed or corrected it.
A Model Context Protocol interface to the full corpus, so that an AI agent can evaluate the architecture directly rather than being handed documents to summarize.
The test result that mattered: deterministic, non-hallucinated outputs. A corpus that produces different answers on different runs cannot be relied on by anyone; one that does not, can be checked by anybody with an agent and no permission from us.
There is no single root. This is one view of the graph, from this node — what you see depends on where you enter and what you are looking for.
Everything above is a distillation. The primary source is published in full, so that any claim on this page can be walked back to what was actually said — and where the two differ, the source governs.
This entry is deliberately incomplete. The gaps below are the things only certain people can settle — and settling one is a contribution like any other: recorded, timestamped and attributed to whoever made it.
Correct anything wrong, add what only you know, or claim this entry as your own. Write to evidence@qpncatalyst.io and Jonathan directly (both are on the button) — a timestamped confirmation comes back with your submission attached, as your own independent record. No account, no permission, no institutional standing required.
Public contributions earn Publicity Premium; bringing others who were there earns Cascade Premium. Everything is valued retrospectively as outcomes verify, so nothing has to be priced up front.
Correct this entry → Add a story, photo or link → Browse the graph →