Exeges: The Grid

A reference in Exeges cites a specific point of a specific version of a document or dataset, the way latitude and longitude anchor a point to the map — this section was amended, this figure is a reporting artifact, this provision is ambiguous.

TL;DR: Working through dense material — legislation, filings, contracts — creates a growing and searchable store of references anchored to the source, addressable, with an identity and lifecycle of its own. References compound and become something you can navigate, query, and audit. A record of what your organization knew, what it decided, and why, all attached to the supporting evidence.

A coordinate system for documents

When a document enters Exeges, every element is given a stable address. Sections, paragraphs, sentences, and table cells can all be cited. A reference is a durable, addressable record that points to one of those addresses. It survives re-rendering, re-indexing, and every change in how the document is displayed.

Because references are addressable, they can cite each other. A newer conclusion can supersede older ones, which stay in the record, retracted but replayable.

Targets and evidence play different roles

An analyst marks a statutory provision as ambiguous. The provision is the target. The justification might be a conflicting committee report, a related amendment, and an older analyst note that argued the opposite — those are the evidence record, and each plays a different role in the judgment. Exeges keeps those roles distinct.

That distinction is what makes references queryable.

• Show every provision marked ambiguous.
• List the findings that rely on this committee report.
• Surface where analysts disagree.
• Track what shifted after a particular amendment.

Each of those is a query over roles.

Three kinds of authors, one layer

Deterministic passes flag whatever you need flagged — every citation, every mention of a name, every dollar figure — anything mechanically identifiable, annotated across the whole corpus at once.

AI agents file claims, and every claim must survive verification. An agent’s finding is checked against the source excerpt it cites before it can enter the record. Quote text that isn’t there, and the whole write is rejected.

People adjudicate and extend. Analysts write notes with typed stances — support, dispute, question, validate — anchored like everything else, including to other annotations. Disagreement isn’t a comment thread; it’s part of the record, and it’s queryable like everything else.

Three author classes, three verification disciplines, one anchored layer.

Compounding value

Ask for every section that adds an amendment and rescinds funding, and the answer is a filtered set: which conditions matched, how many sections, how many occurrences. Start at a clause and traverse to everything that depends on it. Start from an anomaly and walk back to the governing language. Trace how a conclusion strengthened, weakened, or got overturned as new versions arrived — every step of the chain anchored to the text that proves it.

Structured data, same substrate

Quantitative observations — measures, time series, analytic checkpoints — take references the same way: this spike is a reporting artifact; this correlation stops holding after the 2019 rule change. A single reference can span modalities: this statistical trend (structured target) is explained by this clause (document evidence) and contradicted by this prior finding (another reference).

Why it matters

A judgment stored as a chat transcript or a margin note evaporates or drifts away from its source. A judgment stored as a durable annotation stays attached to the passage that justifies it, keeps its history when it’s challenged or superseded, and remains answerable years later to the question every auditor eventually asks: how did you know that?