Every ORBIT screen is designed to leave behind a dated prediction that can be checked later.
The ORBIT Prediction Ledger turns infrastructure analysis into an auditable learning record. Predictions are registered with probability, evidence references, a named resolving source and a due date. Accepted public outcomes are published from the live ORBIT resolution pipeline rather than silently rewriting the original forecast.
History starts before the outcomes are known.
The public ledger launched on October 28, 2026 and contains registered predictions spanning independent public-information cases. Employer-affiliated projects are excluded. Original claims, probabilities, due dates and hashes remain public even if later evidence proves a prediction wrong.
Connecting to accepted-outcome feed…
Current public chain head: c9884cba1c472b58cefd12dda5d9a3cd2eecb564cec17af811961680c9990fd6
Prediction → timestamp → evidence candidate → human review → resolution → calibration.
The public methodology is intentionally simple. Patent-sensitive model mechanics are not disclosed here.
1 · Register
A released screen emits a falsifiable claim, probability, evidence reference, named resolution source and due date.
2 · Observe
Official source changes can generate immutable resolution candidates. A candidate is evidence, not a verdict.
3 · Resolve
Final public outcomes require a separate human review and are appended as confirmed, refuted, partial or void.
4 · Calibrate
Accepted non-void outcomes update public Brier score and directional hit rate. Original forecasts remain unchanged.
Public desk-screen probability bands: 3/3 scenario agreement = 0.85; 2/3 = 0.65; 1/3 = 0.35; 0/3 = 0.10. Lower-confidence claims are retained so the public record does not select only easy forecasts.
No accepted public outcomes yet.
This section is populated only from human-reviewed resolution events tied to public predictions. Candidate evidence and internal review details are never exposed here.
That is intentional. ORBIT will report calibration only after reality supplies a resolvable event and the evidence passes review.
| ID / Site | Constraint | Original probability | Outcome | Observed | Official source |
|---|
Registered before outcomes are known.
These are predictions, not claims of certainty. Each row names the official source used to resolve it later.
| ID / Site | Constraint | Prediction | Probability | Resolution source / due | Status |
|---|---|---|---|---|---|
| Loading public registry… | |||||
Customer predictions are private by default.
Private by default
Customer site predictions are not published without authorization. Public calibration uses only predictions explicitly marked public.
No silent edits
Forecasts, evidence candidates, reviews and resolutions follow append-only records. A correction does not overwrite history.
Bounded public output
The public feed exposes accepted outcomes and aggregate calibration only. Reviewer identity, candidate packets, private predictions and internal rationale remain non-public.
Analysis ends. Evidence accumulates.
The long-term ORBIT asset is a growing record of what was predicted, what evidence supported it, what happened, and how the model should change next.