StatVault

StatVault · Trust pulse

Daily trust health

Trust health at a glance. Receipts measure throughput, canary misses measure honesty, calibration arcs measure user epistemic improvement, and the Integrity Index synthesizes all trust signals into one score.

Efficient by design

How Knox spends →

Most questions are answered for free — from cached evidence, not a paid model call.

58%

answered free ▲ rising

369 / 633 calls

served without model spend

$7.20

projected 30-day model spend

Receipts (7d)
126 issued · top proof-depth 0
126 all-time
Canary misses (7d)
0 misses
no misses on record
Calibration arcs
0 active
no arcs yet
Knox cost
cache delta tracked
Compare crowd calibration
awaiting resolved comparisons (0 pending)
who's-better votes resolve against later SignalRating movement
CSP report sink
fs backend · 0 reports (7d)
⚠ filesystem fallback in production — provision STATVAULT_REDIS_URL
Low-confidence refusals (7d)
0 refusals
no refusals on record
Knox Integrity Index
77.1 / 100 · grade B
limiting: verification (20%) · 1 witness pubkey(s) · 1d old
Per-sport integrity
1 sport
nfl
31.1 F

Refreshed via npm run audit:sport-integrity-index

Conviction distribution
4 topics · median 1.00
4 stable 0 mixed 0 contested

How often Knox holds a steady read. Refreshed via npm run audit:conviction-distribution.

Self-tuning system readiness
Confidence bandsAWAITING
No resolved-outcome corpus on disk; bands on static fallback
npm run snapshot:compare-pair-outcomes && npm run build:proxy-corpus
Compare calibrationAWAITING
No resolved compare pairs yet
npm run snapshot:compare-pair-outcomes
Micro-vote corpusAWAITING
0 votes — the self-tuning loop is unfueled
Vote on the homepage 'boldest Knox call' spotlight
Witness verificationLIVE
1 witness pubkey(s)
Contradiction rate (14d)
0.00% · 0 of 251 answered
no contradictions on record
Self-critique calibration
3 rules active
9 total fires · no resolved outcomes yet

Generated 2026-05-21T17:05:23.056Z