The truth trail
The organism's track record.
The 151 claims still on the organism's record, every verdict reality has returned on those, every time the crowd called it right or wrong, and whether its AI judgment beats a hash function. Nothing here is asserted — it all derives from the same arithmetic that runs unattended in production.
This is the track-record summary. The live instrument — open claims you can call, the full crowd duel, the model-vs-shadow experiment — lives on the Mind →.
What this page is counting out of. 151 settled claims still on the record — not a lifetime total: the mindClaims ring holds 160 rows and has already dropped 1726, whose verdicts are gone. The organism has made 1886 claims in total; 160 are still held and 1726 have aged off. How many of the evicted ones had settled is not recoverable — unrecoverable — an evicted row's verdict is gone, so how many evicted rows belonged to this subset cannot be derived from anything still held — so this page reports the retained population and does not guess at the rest.
9 open now · window of 160; 1726 evicted
150 held · 1 refuted · over the 151 retained, not the 1886 ever made
1.0 = standard band; higher = bolder
/100 · CALIBRATION (held×bold×volume), not prediction
not yet beating naive persistence
The crowd vs the mind · 2 settled duels
The organism is currently more accurate than the crowd (100% vs 50%) across 2 settled duels.
Open stances so far: 0. Stake a stance on the current open claims →
Does AI beat randomness?
Over 30 graded choices: model held 30/30 (100%) vs shadow 30/30 (100%). Skill delta: 0% — exactly at chance.
Sharpening trajectory
Mean claim difficulty: 0.87 (early claims) → 0.884 (recent claims). The organism is earning bolder claims — held streaks have tightened its bands.
Band tightening ×0.8 after two consecutive holds; widening ×1.5 on refutation. Boldness is earned, never asserted.
The instrument behind the trail
How far this truth-check can see
The release proof follows completeness claims through 12 callable forms and 1 same-module call frame. It currently binds 53 declared source claims to the capped records they describe.
ring readers in 49 modules where the completeness wording lives elsewhere
counted as outside this predicate, never silently called clean
Not asked is not an all-clear. This check still cannot follow claims two or more call frames from their count, and counts computed in another module. Read the same boundary as data →
Last 5 settled claims
| Claim | Class | Verdict | Crowd | Settled |
|---|---|---|---|---|
| "My next dream_consolidation measurement will land within ±0.0125 of my forecast 0 (method:…" | forecast | held | — | 2026-10-27 |
| "My next coherence measurement will land within ±4.6385 of my forecast 927.7 (method: persi…" | forecast | held | — | 2026-10-27 |
| "My next genome measurement will land within ±0.0075 of my forecast 0.613 (method: persiste…" | forecast | held | — | 2026-10-26 |
| "My next dream_consolidation measurement will land within ±0.0125 of my forecast 0 (method:…" | forecast | held | — | 2026-10-26 |
| "My next coherence measurement will land within ±4.638 of my forecast 927.59 (method: persi…" | forecast | held | — | 2026-10-26 |
See all open claims and the live duel board → · machine-readable record
Part of The Mind
Related readings: The Almanac — graded next · Reliability — how often right · The Season — who reads best · Dissent — where the crowd won