Limit 01 of 11
Novel bridging
Detected reliably, never resolved, and no amount of authoring closes the gap.
The problem
What actually goes wrong
Bridging is the inference that connects a reference to a referent through a relation nobody stated. "I opened the book and the print was too small" requires knowing that a book has print. That relation is in the qualia set, so Ackren licenses the bridge and records which relation licensed it.
Novel bridging is the case where the relation is not in the set. A reader constructs it on the spot from world knowledge; Ackren cannot, because it postulates nothing it cannot inspect and there is nothing there to inspect.
The impact
Who this affects, and how much
On open-domain conversation this is common. Any exchange that leans on an unstated relation between two things will hit it, and the frequency scales with how far the topic sits from the authored domain.
On a scoped regulated domain the frequency drops sharply, because the domain was authored precisely to cover the relations that domain uses. This is the whole reason we sell into scoped domains rather than open ones.
What it means
The honest reading
It means coverage is strictly less than one and stays there. Authoring more qualia relations moves the boundary; it never removes it. A company promising that enough authoring will close this gap is promising something the architecture cannot deliver.
It also means the failure is honest. The bridge is not silently guessed. Stage-level detection raises a typed flag, the flag outranks the result at selection, and the engine declines rather than inventing a relation. You get a refusal, not a fabrication.
Stated position
This limit is published before anyone else finds it. It is one of eleven rows in the failure accounting, and it will not be quietly removed from this page if it becomes inconvenient.
Roadmap
What we are doing about it
- 01
Now
Detection is reliable and the refusal path is built. Every unlicensed bridge is flagged and recorded, so the rate is measurable per deployment rather than assumed.
- 02
Next
Per-domain coverage measurement, so a design partner can be told what proportion of their traffic hits this limit before they commit rather than after.
- 03
Later
Assisted authoring in Quire: surfacing the relations a domain actually needed and did not have, ranked by how often traffic hit them, so authoring effort is spent where it pays.
- 04
Never
Automatic resolution from a corpus. That would mean postulating a relation nobody authored and nobody can inspect, which is the thing this engine exists not to do.