Platform prototype

The TargetSpace loop, as a working system

This is a functioning prototype of the evaluation layer behind the TargetSpace benchmark: passive longitudinal evidence becomes sealed prediction instances; model predictions are hashed and sealed before outcomes exist; deterministic rules resolve outcomes; and a scoring pipeline reports calibrated, target-specific skill against R1/R2 baselines, controls, and evidence costs.

Prototype · All data on these pages is synthetic (generated personas and machines; no real personal data). The registry runs locally in your browser. Numbers illustrate mechanics, not empirical results.
How it works

Observe → represent → predict → calibrate → minimize evidence → validate

Observepassive longitudinal evidence
Representmodel builds target state
Predictsealed distribution over next states
Calibrateprobabilities match frequencies
Minimize evidencelift per cost axis
ValidateR1/R2, wrong-target, gates

A system earns target-specific credit only when its sealed predictions beat the population prior (R1) and the target's own routine (R2), stay calibrated, collapse under wrong-target permutation, and disclose what the evidence cost.

Walk the loop

Product surfaces

1 · Admin

Create benchmark run

Define target type, evidence streams, horizon, answer spaces, resolution rules, baselines, and cost axes.

2 · Admin

Instance constructor

Watch passive timestamped evidence become sealed prediction instances with cutoffs and deterministic rules.

3 · Lab

Submit predictions

Submit a probability distribution over the operational possibility space — the answer states still open; it is hashed and sealed before resolution.

4 · Registry

Sealed prediction ledger

Tamper-evident registry: hashes, timestamps, statuses, audit trail, and chain verification.

5 · Resolution

Outcomes

Deterministic rules resolve future evidence into outcomes — never self-report, never post-hoc judgment.

6 · Scoring

Scoring report

Log skill in bits vs R1/R2, Brier, calibration, wrong-target control, ablation, and validity-stack gates.

7 · Efficiency

Evidence efficiency

Incremental lift per evidence-cost axis: which observation earns its cost, and which does not — toward the minimum sufficient observation.

8 · Privacy

Federated mode

Architecture for keeping raw personal data local: only sealed predictions, outcomes, and aggregates leave.

Schemas

Data model

Twelve JSON schemas define every record: runs, instances, sealed predictions, outcomes, reports.

Honest scope

What is implemented vs conceptual

ComponentStatusNotes
Sealed prediction registry — tamper-evident ledger (hash-chained)implementedLocal (browser storage); production would add an external timestamp witness and database
Instance constructor from synthetic evidenceimplementedDeterministic rules pre-registered per instance
Scoring pipeline (log skill, Brier, R1/R2, controls)implementedPure functions; runs end-to-end in the browser and in CI
Evidence-efficiency reporting (per-axis)implementedOrdinal 0–3 cost axes in the prototype; measured units in production
Validity-stack gates & certification-style summaryimplementedSummary of gate outcomes only; no certification marks exist
Live model inferenceconceptualDemo distributions are hand-authored synthetic submissions
Federated local runner & encrypted audit packageconceptualArchitecture specified on the federated page
External timestamp witnessconceptualPrototype witness is the local ledger chain itself
One click runs the whole loop: