DAEPOLIS
Judgment infrastructure
AboutMethodBriefsLedgerSurfaceCohort
DAEPOLIS

Decision records for institutional judgement.

DAEPOLIS turns high-stakes analysis into navigable intelligence surfaces: claims, evidence, dependencies, risks, forecasts, monitoring triggers, and decision records. The product is the record object; prose is only one readable view.

Know · judge · decide · monitor · revise · learn
Public RegisterLive
Active Commitments
—
Resolved
—
Scored Forecasts
—
Validation Window
Q3 2026 – Q4 2027
Commitment-hashed · Brier-scoredOpen Public Register →
What DAEPOLIS is

Institutional judgement infrastructure for durable records.

01
Decision Record
Question, decision owner, options, chosen action, approvals, monitoring triggers, and later outcome.
02
Claim Trace Audit
Existing prose is decomposed into claims, warrants, belief debt, support state, missing counter-evidence, falsifiers, and questions before reliance.
03
Claim Ledger
Canonical claims, support state, evidence links, forecasts, Brier scoring, disputes, and retractions.
04
Source Hardening
Provisional, source-assisted, source-grounded, or contradicted states with provenance and gaps visible.
05
Review Workflow
Human challenge, approval, hardening tasks, source requests, and contestable record updates.
06
Forecast & Calibration
Probability distributions, reference classes, red-team sidecars, decisive unknowns, Brier scoring, and calibration memory.
07
Surface Instrument
Trust capsules, role-specific views, append-only surface events, signal intake, and node-level proof chains.
08
Exports/API
Narrative PDFs, record packages, audit-trail exports, public proof paths, and enterprise data surfaces.
System Object Spine

The frontend should expose the backend spine.

DAEPOLIS is no longer just an analysis form. The backend now produces record objects: ingestion reports, source states, claim ledgers, prediction distributions, decisive unknowns, review events, outcome memory, trust capsules, and Surface deltas. The product surface has to make those objects visible, inspectable, and usable.

The quality rule is simple: if the backend can prove, warn, score, repair, or remember something, the frontend should not hide it behind a generic text box.

01Intake Integrity

Documents enter as untrusted evidence, not authority.

PDF/Office extraction integrityUntrusted-content boundarySource packsHardening tasks
Show extraction quality, provenance, suspicious content, and review work before any audit treats the document as usable evidence.
02Claim And Evidence Core

Claims are separated from prose and kept provisional until supported.

Canonical claim ledgerSource-hardening statesSemantic collisionsEvidence gaps
Expose the support state, missing source class, contradiction path, and next hardening action beside the claim itself.
03Judgement Memory

A record keeps learning after the first answer.

Decision RecordsAppend-only eventsReview workflowOutcome ledger
Let users see who changed what, why the record moved, what remains disputed, and what later happened.
04Forecast And Calibration

Probability is a public object, not a sentence.

Prediction distributionsReference classesRed-team sidecarsCalibration surface
Show the distribution, outside-view class, adversarial disagreement, decisive unknowns, and update history instead of only one midpoint.
05Monitoring And Repair

The system should wake when reality changes.

Decisive unknownsMonitoring triggersLineage decayError memory
Turn uncertainty into watchable signals: what would change the judgement, what has gone stale, and what prior mistake should be reused.
06Surface Instrument

The graph is an operating surface over the record spine.

Trust capsulesRole-specific viewsSurface eventsSignal intake
Every node should answer why it matters, why it can be trusted, what is missing, what changed, and what role should do next.
CJI Validation

Testing the axis conventional indices miss.

The CJI validation surface separates orthogonal stress measurement, retrospective method probes, and sealed future predictions. Directional tests are live; measured-panel country findings are not claimed yet.

Open CJI Validation →
Orthogonality
CJI carries stress information not captured by formal-democracy and prosperity proxies.
Retrospectives
Historical probes test limits, ordering, matched pairs, negative controls, and method bugs.
Pre-registration
Sealed future calls are the predictive-credit surface; hindsight cannot touch them.
Surface Instrument

The Surface should become a living institutional instrument.

The next foundation is controlled access: let users feed the Personal Intelligence workspace with new signals, let lawful monitors find deltas, and make every update land as inspectable memory. The graph should not merely look good; it should tell the user why a node deserves trust and what would change it.

Manual signal intake
Personal Intelligence should accept new observations, source snippets, board notes, and monitoring updates directly inside the workspace. They become logged evidence packets with owner, timestamp, source posture, affected node, and review state rather than loose notes.
Autonomous delta monitoring
Official registers, public APIs, RSS/news feeds, licensed datasets, user-uploaded source packs, and scheduled watchlists should wake only the affected claims, dependencies, triggers, and Decision Records when something material changes.
Database-backed memory
A surface is strongest when new briefs, imported documents, signal updates, verification-chain edits, and monitoring deltas land in Supabase as durable records. Static samples are useful demonstrations; production judgement needs freshness and migration hygiene.
Node-level trust chain
Every object on the Surface should answer: why should I trust this node, what evidence supports it, what is missing, what changed, what is anchored, and what action would harden it next.
Launch Workbench

Pick the right intelligence job

Method

The job picker creates a record shell. Method routing, source hardening, claim extraction, review status, and ledger objects run underneath.

The record spine.

01
01
Question
Decision owner, institutional context, public-good exposure, horizon, and evidence-that-would-change-mind.
02
02
Evidence
Sources, excerpts, hashes, proximity, reliability, contradictions, and hardening state.
03
03
Claims
Canonical claims, candidate entities, relationships, forecasts, falsifiers, and monitoring triggers.
04
04
Decision
Review status, dissent, disputes, exports, later resolutions, and calibration/error memory.

The same spine supports Judgement Surfaces, ledger entries, review workflow, narrative exports, and later dispute or retraction history.

IR
Intelligence Record Core
The central object is the structured record behind a judgement: question, entities, evidence, claims, dependencies, review, decision, monitoring, and outcome.
AT
Audit Trail First
Material events become inspectable history: claim creation, source changes, probability updates, reviews, disputes, and exports.
QA
Weak Signals Marked
Quality gates surface provisional evidence, weak rhetoric, unsupported document claims, unresolved decisive unknowns, and model-fit rejection rather than hiding them.
Current Product Surfaces

The code already stores more than prose.

IRCore Object
Institutional Intelligence Record

The institutional container behind the judgement: question, decision owner, entities, source pack, claim ledger, evidence graph, dependencies, forecasts, dissent, approvals, monitoring triggers, and outcome.

Decision contextSurfaceOutcome
CTBrief Forensics
Claim Trace Audit

Existing prose is reconstructed into a belief architecture: material claims, warrants, hidden priors, belief debt, missing base rates, counter-evidence gaps, falsifiers, and board-ready questions before reliance.

ClaimsWarrantsBelief debt
DMInstitutional Dependency
Materiality Analysis

Risk-bearing relationships are separated into dependency type, materiality, evidence state, reputational channel, business risk, legal/regulatory risk, rhetorical downgrade, triggers, and next actions.

DependenciesMaterialityDecoder
CLLedger Memory
Canonical Claim Ledger

Extracted claims remain provisional until reviewed. Support status, evidence chain, prediction resolution, disputes, and retractions stay visible across analyses.

Claim stateEvidence chainRetractions
RHHardening Layer
Source and Review Workflow

Source-hardening states, review queue, hardening tasks, rhetoric gates, and adversarial checks make the system willing to mark its own weak analysis.

Source stateReviewRepair
Institutional Decision Infrastructure

Can an institution defend how it knew, judged, decided, monitored, revised, and learned?

That is the product test. DAEPOLIS is being built around the record, not the rhetorical finish: claims, evidence, dissent, forecasts, review status, monitoring triggers, and later outcomes in one inspectable object.

Responsible Capital Records
Portfolio exposure, end-use risk, sovereign concentration, issuer dependency, evidence gaps, and named review triggers in one record.
Public-Power Accountability
Procurement, platform, legal, technology, and institutional-capture cases translated into claims, sources, actors, and monitorable tests.
Judgement Memory
Surfaces, calibration, claim versions, source degradation, disputes, and later outcomes show how the system learned or failed.
Access

Early-stage design partner cohort opens in autumn 2026.

The public site carries the credibility surface that matters at this stage: LSE affiliation, MSc-level method documentation, the canonical-claim ledger, published briefs, and a visible calibration/error record. Institutional names remain private unless explicit Stage 2 permission is granted.

Public Surface
Available now
Open method, public ledger, and published briefs.
◆ Canonical-claim ledger and prediction register
◆ Public methodology documentation
◆ Published intelligence records and narrative exports with sensitive recipients redacted where relevant
◆ LSE affiliation and MSc-level political-economy method surface
Recommended
Design Partner Cohort
Autumn 2026 launch
For teams willing to test judgement infrastructure on real cases.
◆ Structured decision records rather than standalone reports
◆ Claim/evidence review workflow and audit-trail exports
◆ Forecast commitments with resolution criteria
◆ Private feedback loop on trustee-defensibility Judgement Surfaces
Method Partnership
Permissioned
For organisations that need controlled methodology documentation.
◆ Versioned method packs and governance workflow
◆ Framework validations, disputes, and retractions
◆ Source-hardening and provenance requirements
◆ No public institutional naming before Stage 2 permission
Infrastructure API
Staged
For later integration into GRC, ESG, stewardship, and board workflows.
◆ Claims, evidence, predictions, and resolutions as data
◆ Power ontology and framework-validation endpoints
◆ Portfolio and holdings ingestion once ownership controls are complete
◆ External anchoring and signed exports before institutional launch