DAEPOLIS is institutional intelligence.

Reconstructs the claims, warrants, evidence, uncertainty and belief debt inside material an institution is being asked to rely on.

Turns warranted expectations into resolvable commitments with distributions, falsifiers, later outcomes and calibration memory.

Measures contribution, agency and lived outcomes across groups and time while keeping measurement validity separate from causal and predictive claims.

Connects reliance, forecasts and CJI through actors, arrangements, capabilities, alternatives, affected groups, outcomes and falsifiers.

Governance after reliance: durable records containing the elected decision, evidence state, objections, monitoring triggers, review authority and later outcome.

PDF and Office uploads must produce trusted extraction reports before the text can enter the untrusted-content boundary. Raw private files are not stored.

Existing consultant reports, AI memos, ESG briefs, policy papers, board papers, and investment notes can be decomposed into claims, warrants, implied priors, belief debt, missing counter-evidence, falsifiers, adversarial objections, and questions before reliance.

Companies, countries, buyer markets, sources, claims, contracts, portfolios, regimes, public agencies, vendors, and decision records can be audited as risk-bearing dependencies with separate materiality channels, evidence gaps, monitoring triggers, rhetorical downgrades, and next actions.

The post-login Surface is becoming the live instrument: users should be able to add observations, source snippets, reviewer objections, and monitoring deltas to specific nodes while the system preserves provenance, freshness, review state, and the verification chain.

Atomic claims with support status, evidence links, verification state, disputes, retractions, and resolved forecast outcomes.

Evidence states distinguish provisional, source-assisted, source-grounded, contradicted, and missing-source material.

Candidate claims, evidence, and hardening tasks can be challenged, approved, rejected, or returned for source work.

Predictions carry probability distributions, reference-class context, red-team distribution sidecars, decisive unknowns, resolution criteria, Brier score, and calibration bins.

Trust capsules, append-only Surface events, role-specific record views, and node-level proof chains make the graph inspectable rather than decorative.

Brief PDFs, production packages, audit-trail exports, proof downloads, and enterprise data resources expose the same record spine.