Skip to main content

Trustworthy by default.
Nothing hidden.

The discipline behind our research goal — the Verifiable Knowledge Foundation: how approved inputs become a scoped knowledge foundation, how AI stays grounded in traceable evidence, and how consequential decisions stay visible to people. DjeedX, Anchor, and Gateway use different surfaces, but the same rules apply: provenance, isolation, explicit permissions, cite-or-abstain behavior, and measurable evaluation.

ProvenanceBerkeley Protocol-readySigned receiptsBronzeSilverGold

One record's journey

From seven fragments to one defensible decision.

The same event reaches you as five articles and two reports. This is how Djeed holds it once, connects it, and hands you something you can inspect and defend. The principles below are the guarantees behind each step.

01 · FRAGMENT

The same event, seven times

Five articles and two reports describe one incident: different words, different details, no shared identity.

02 · SHAPE

Held once, with source spans

Djeed holds the event as a single record: what, where, when, who. Every field keeps the exact passage it came from.

03 · CONNECT

Linked to what it touches

The record connects to the actors, places, and related events it mentions, so one question can follow the connections across records.

04 · INSPECT

Four lenses, one record

Table, graph, map, and the provenance panel are views of the same record, not four copies. This is the DjeedX working surface.

Explore DjeedX →

05 · DECIDE

A decision you can defend

Export a brief with every claim traceable to source, or hand the judgment call to a person. Confidence is stated, never implied.

Try Gateway →Deployment boundaries →

Principle 01Provenance on every record.

Every record promoted into a Djeed knowledge foundation keeps the evidence needed to inspect it: origin, capture time, transformation history, confidence, and review state.

AI-extracted records also store the upstream extraction payload — the raw response from the model, not just our parsed interpretation — so any disputed value can be re-checked against what the model actually returned.

provenance bundle
source_urlhttps://…/article/123
captured_at2026-04-26T08:14:02Z
content_hashsha256:a4f5…
extracted_byagent v3.2
edits3 (last 2026-04-25)

Principle 02Evidence-ready outputs.

Djeed's evidence discipline draws on established digital open-source investigation practice, including the Berkeley Protocol: collect, preserve, verify, and present material so the work can be inspected under scrutiny. The principles travel across sectors. The same discipline shows up whether the work is tracking permit pipelines across municipalities, watching industrial emissions disclosures against open environmental data, following large-project financing flows through procurement portals, or building entity graphs from corporate registries for a parcel or asset — and DjeedX connects that discipline to your own internal records.

Two things make structured intelligence at scale possible without losing the discipline, and Djeed combines them deliberately. AI extraction reads the open record at machine-supported scale — turning unstructured prose into typed claims, line by line, source by source. That is the GenAI side: a reading machine. The Berkeley-aligned methodology is not a layer that sits on top of it — it is embedded by construction. The original source URL, capture timestamp, content hash, raw extraction payload, and corroboration trail are captured at the moment of reading, which is what makes the methodology and the extraction inseparable. The OSINT methodology preserves the chain of custody; AI makes it feasible at the scale the modern public record actually requires. The investigator's judgement stays where it belongs — human. The mechanical reading does not.

The same principle applies beyond extraction. A general-purpose agent is not a methodology, and one organisational chatbot is not an operating model. Effective implementation begins with the people who understand the work: mapping the knowledge, redesigning the workflow, and encoding a repeatable sequence in which specialist agents, deterministic operations, and accountable human decisions each have a defined place.

PRINCIPLE

Provenance captured at the source

PRINCIPLE

Chain of custody preserved end-to-end

PRINCIPLE

Corroboration across independent sources

Djeed provides the evidence trail and operating guardrails. Your accountable people still set the source bar, review exceptions, and make the consequential decision.

Every Silver-tier record carries the chain-of-custody attributes the Protocol expects: original source URL, capture timestamp, content hash, raw extraction payload, deduplication trail, and version history. The analyst — you — runs the verification, source-assessment, corroboration, and review steps the methodology calls for.

We power the workflow with products and tools — DjeedX, Anchor, Gateway, and the evidence trail — that make disciplined research possible at the speed and scale your work actually demands: whether a planning team is mapping zoning changes against developer filings, an environmental researcher is comparing corporate commitments to what filings actually disclose, or a due-diligence analyst is tracing ownership, litigation, and permit history across registries.

Principle 03We transform into Silver. Gold is yours to build.

The Djeed method restructures approved inputs through three layers:

BRONZESILVERGOLD
Bronze· Raw input

Your files, the open record, social-media signal — direct extraction with the original URL preserved. Internal layer; never surfaced as a finished dataset.

Silver· The trusted working layer

Structured, deduplicated, entity-resolved, evidence-linked, and ready for retrieval or analysis inside the surface that owns the foundation.

Gold· What you build

Verified in your context. Combine Silver with observations, policies, expert judgment, and an explicit decision or operating workflow.

The handoff is intentional. Djeed creates a traceable working layer; accountable people apply context, approve consequential actions, and own the result. Agents may assist both sides, but they do not erase the boundary.

Principle 04

Approved inputs stay tied to their sources.

Public sources, customer records, approved folders, and system inputs come through the same Bronze→Silver pipeline that restructures your own inputs, with explicit evidence discipline. Each record keeps a citation edge to its origin. Edits, interpretations, and decisions stay inside the foundation that owns them; the upstream source is not silently mutated. That's the path from Silver to your own Gold.

Principle 05

Isolation before connection.

Every customer, app, profile, and deployment has its own knowledge foundation and tool registry. Nothing is matched or shared across scopes by default. Any exchange must be explicit, logged, purpose-limited, and revocable.

Principle 06What documents a knowledge foundation.

A governed foundation needs a receipt: what scope it covers, which inputs were approved, how records are structured, what transformations ran, which evaluation gates apply, and where people review or approve. Public methodology explains the shared discipline; each customer keeps the implementation receipt for its own scope.

Coring (unit of observation)

Every structured collection declares its primary unit — what one record represents. A foundation can be cored at one of:

  • Event-centric — one row = one event (incident, decision, action), with linked actors, locations, dates, and sources.
  • Claim-centric — one row = one assertion by one source about a fact in the world. Atomic, never merged across sources.
  • Act-centric — one row = one action (often nested inside an event); granular for deep operational tracing.
  • Entity-centric — one row = one organisation, person, group, or asset, aggregated across all its appearances.
  • Indicator-centric — one row = one measure at one place at one time (statistical surface).

The choice depends on the workflow and the decision it supports. The foundation receipt declares the unit up front and ties every record back to that decision.

Bronze → Silver migration

Records start at Bronze (raw extraction with the source URL preserved). They become Silver when they pass the migration gate. The pipeline runs:

  1. URL dedup — already-processed sources are skipped before any extraction work happens.
  2. Spatial enrichment — lat / lon is reverse-geocoded to authoritative country / admin1 / admin2 codes via shapefile join, plus an H3 hex cell (resolution 7) for fast spatial blocking.
  3. Entity resolution — fuzzy name matching collapses spelling variants of the same actor across sources.
  4. Cross-source corroboration — multiple claims about the same event are linked together, with confidence scored from source reliability + claim agreement.
  5. Promotion gate — thresholds for confidence, corroboration, completeness, and human review are configured for the workflow. Material below the gate remains Bronze and cannot be presented as trusted output.
Deduplication scoring

At the dedup stage, every candidate pair is scored across five dimensions:

  • Spatial proximity — same H3 cell or within km distance threshold.
  • Temporal proximity — same day, or within N days when date precision is fuzzy.
  • Semantic similarity — claim-text embeddings cosine similarity above threshold.
  • Category match — primary + secondary action categories must align.
  • Entity match — actor / victim names resolve to the same canonical entity.

Pairs above the auto-merge threshold (typically same-day, within 5 km, same category) merge automatically. Borderline pairs go to AI review; hard cases escalate to human review. Every merge keeps a trail back to the source records — provenance is never lost.

Silver working layer

A user or agent working inside the owning scope sees a Silver layer with:

  • Typed rows following the coring (one event / claim / act / entity / indicator per row).
  • Per-record provenance bundle — source URLs, capture timestamp, content hash, raw extraction payload, full edit history.
  • Spatial fields — lat / lon, country, admin1, admin2, H3 cell.
  • Temporal fields — event date, date precision, week / month / year buckets.
  • Confidence score per record.
  • Verification status field (unverified / verified / disputed).
  • Linked entities, events, sources — explorable in the product surface that owns the foundation.

Available views and export formats depend on the product, permissions, and the data contract for that deployment. An export never widens the owning scope.

RoadmapOpen questions we're still working on.

Per-record confidence-decay over time for AI-extracted records.

Public methodology versioning — every breaking schema change tagged with a migration note + diff endpoint.

Independent third-party methodology audits on the highest-evidence datasets.

A formal Silver-to-Gold playbook — patterns for how teams cook Djeed Silver into decision-ready Gold inside their own workflow.

Per-dataset Berkeley-Protocol coverage matrix — for each Silver dataset, which methodology attributes are guaranteed, which are best-effort, which are out of scope.

Built to be inspected. Trusted by default.

See the method inside DjeedX, apply it through Anchor and Gateway, or scope a workflow with an explicit evidence and measurement plan.

See pricingScope a workflow