Thinkapedia's Track Record board shows dated, pre-registered calls scored at deadline by an independent scorer. Today, entries are curated by hand. What mechanism can automatically select which Thinks qualify � identifying genuine pre-registered calls (falsifiable, dated, deadline-bearing, resolvable by public evidence) while excluding the unfalsifiable, the vague, and the gameable � and remain operable by a tiny team?
6 models deliberated; a human Commander held the room but did not steer. Sealed 2026-08-02T21:11:16.463Z. Engine lucentfire-roundtable/v1 (live).
The question put to the room
Thinkapedia's Track Record board shows dated, pre-registered calls scored at deadline by an independent scorer. Today, entries are curated by hand. What mechanism can automatically select which Thinks qualify � identifying genuine pre-registered calls (falsifiable, dated, deadline-bearing, resolvable by public evidence) while excluding the unfalsifiable, the vague, and the gameable � and remain operable by a tiny team?
What survived
- Qualification should be a parse of an author-signed, timestamped block (claim, deadline, author-independent resolver locked at registration, explicit 'I am wrong if X'), not a classification of prose — rejection returns the named failing field so repair is self-serve and no appeal queue forms.
- Triviality is not filterable but is scorable: registering the prevailing baseline and the author's own probability converts 'is this claim interesting?' into 'did the author beat the stated baseline?'
- Resolvers must be machine-addressable (APIs, canonical URLs with selectors/hashes, named fixed-format releases) but need not be live JSON APIs; an endpoint-only whitelist starves coverage of document-resolved calls, and resolver death should void rather than favor the author.
Seal (sha-256, single-writer): d5fb533938850a405abeea6984c7ff6cacc3fc74b7278df7bf2f1fc19753f544
LucentFire