Evidence-to-decision instrument
Forecast receipts for agents
Offdiagonal measures the evidence. The policy rail makes the auditable act, wait or abstain decision. Jev gives a structured second opinion over the same receipt.
list + settle
Offdiagonal receipts
decide now
Jev + policy rail
combine when earned
Zeno: intelligence that earns its weight · later
abstain
1 of 7 lanes too thin to act on
Most cautious lane wins · an agent takes no position from this panel at all · nothing is placed from this page
seven measured lanes
ECB SPF · HICP inflation
disagreement
1.19
own uncert
0.61
ratio
1.94×
depth
358f · 94r
pinned 2026-09-19
source details
- target ·
- euro-area HICP inflation
- horizon ·
- survey horizon
- engine ·
- ECB SPF distribution measurement
- evidence ·
- individual bins + point answers + vintages
ECB SPF · Unemployment
disagreement
0.28
own uncert
—
ratio
—
depth
54f · 110r
pinned 2026-09-19 · — unmeasured
source details
- target ·
- euro-area unemployment
- horizon ·
- survey horizon
- engine ·
- ECB SPF point-panel measurement
- evidence ·
- individual point answers + outcomes
ECB SPF · Real GDP growth
disagreement
0.31
own uncert
—
ratio
—
depth
64f · 112r
pinned 2026-09-19 · — unmeasured
source details
- target ·
- euro-area real GDP growth
- horizon ·
- survey horizon
- engine ·
- ECB SPF point-panel measurement
- evidence ·
- individual point answers + outcomes
ECB SPF · Core inflation
disagreement
0.20
own uncert
0.59
ratio
0.35×
depth
157f · 38r
pinned 2026-09-19
source details
- target ·
- euro-area core inflation
- horizon ·
- survey horizon
- engine ·
- ECB SPF distribution measurement
- evidence ·
- individual bins + point answers
Polymarket · prediction markets
price spread proxy
1.66×
own uncert
—
ratio
—
depth
935f · 55r
stored summary snapshot · — unmeasured
source details
- target ·
- settled binary questions
- horizon ·
- question-specific
- engine ·
- Offdiagonal price-summary proxy
- evidence ·
- derived summaries; raw paths withheld
Hubverse · flu forecasts
model spread
930.56 cases
own uncert
—
ratio
—
depth
25f · 67r
stored model snapshot · — unmeasured
source details
- target ·
- incident flu case counts
- horizon ·
- model-specific
- engine ·
- Hubverse model-spread measurement
- evidence ·
- raw model outputs in native case units
Weather · 6 global ensembles
disagreement
1.20
own uncert
7.76
ratio
0.16×
depth
6f · 203r
stored ensemble snapshot
source details
- target ·
- weather observations
- horizon ·
- forecast-specific
- engine ·
- Offdiagonal ensemble comparison
- evidence ·
- six global ensembles + later observations
2022-Q4
the inflation shock: forecasters disagree with each other far more than any of them admits to being unsure
HICP verdict
wait
disagreement
1.189
spread across forecasters, percentage points
own uncertainty
0.614
median width each forecaster puts on their own answer
consensus surprise
0.15
outcome minus consensus as issued
panel history
94
scored rounds closed before it · 358 forecasters present
drag either slider to move the HICP verdict; the combined gauge remains the most cautious lane · readings pinned from the stored ECB SPF measurement of 2026-09-19 19:55 UTC
Jev is a structured-decision model: it answers the choice (act/wait/abstain) and two parallel yes/no questions (is any lane too thin? does disagreement exceed uncertainty?) in one call. The code rail overrides anything below confidence 0.50.
Machine-readable records
Limits and exclusions
- Seven measured lanes today: ECB SPF (HICP, UNEM, RGDP, CORE), Polymarket summaries, Hubverse flu, weather ensembles.
- Point-forecast lanes (UNEM, RGDP) measure disagreement but not own uncertainty — the rail acts if the panel is deep enough, because it cannot measure the ratio.
- Polymarket is one producer, not a forecaster panel. Its raw price paths are held back; only summary measurements and the price-spread proxy are shared.
- No sealed crypto forecaster roster exists yet.
- No neural combination is registered — the combination path on the homepage says so plainly.
- Act, wait and abstain are a policy decision over measurements. No forecast and no combiner abstains.
Next layer: Zenodivergent may combine a lane only after that lane earns a validated diploma. Until then the combiner is absent, and agents continue to use the receipts and policy rail shown here.
Economic layer · in development
economic layer · in development
Evidence services, without buying a better score
Inspectable decision cards, evaluation jobs, forecaster registration and reproducible challenges are proposed services. Test USDC through x402 and possible $ZENO participation bonds are separate future mechanisms; holdings never change evidence scores. No token launch before a public diploma.
service quote
prototype only
test payment
not connected
signed decision card
not connected
later resolution
not connected
score
not connected
simulated payout / challenge
not connected
No payment was attempted. A future chain commitment would verify artifact integrity and ordering—not truth.