Under constructionThis deployment is live for testing only. Data, features, and findings may change without notice. To contribute — code, tips, corrections, or legal review — info@lexdoge.org.
SourcesBudget PDF· 2h|Legistar· 18m|Open Data· 47m|KPPA Pension· 4d|EMMA Bonds· 1dBuildsha256:dev·Next ingest: 23m
ABOUT

What LexDOGE is — and what it is not

Mission

LexDOGE is non-partisan. LexDOGE is not a political organization.

LexDOGE exists as a community service: an open-source, freely accessible window into how public officials and public institutions in Lexington-Fayette allocate public capital — and whether those allocations are responsible, consistent with the public record, efficient, and in conformance with the laws of the community.

The federal DOGE dashboard is the cautionary template — it claimed $214B in savings while only $1.4B was independently verifiable. LexDOGE's design principle is the inverse: flag, never accuse; data integrity over spectacle. Every published claim cites a source document, page, and section. Every AI-generated paragraph carries a confidence score and passes through tiered human review before anything characterizing a named individual reaches the public surface.

Independence

LexDOGE is not an activist organization and not a reactionary one.

It does not endorse candidates, parties, factions, or policy positions. It does not draw conclusions about motive, intent, or character. It does not align with, endorse, or coordinate the projects of any other organization — civic, journalistic, academic, governmental, or otherwise — and it does not enter into collaborative campaigns that would compromise editorial independence.

It interfaces directly and unmediated with the community it serves. What it publishes is a function of what the public record contains and what well-accepted quantitative methods surface from it — not a function of which institutions are willing to amplify it.

What you read here is what the data says. Decisions about what it means — politically, legally, or morally — belong to citizens, journalists, council members, auditors, and the courts. Not to LexDOGE, and not to the model.

Direct community interface

Anyone in the community can contribute. Data submissions, tips, document leads, source-document corrections, FOIA suggestions, and code contributions are welcome — anonymously, pseudonymously, or under your real name, your choice. There is no application process, no membership, no political litmus test.

Pseudonymous contribution is a first-class pattern, not a tolerated edge case. Civic transparency work that relies on insider knowledge of how public institutions actually operate often depends on contributors whose identity must remain protected. LexDOGE's contribution surface is designed for that reality.

Data, tips, document leads
info@lexdoge.org
Email. Pseudonymous senders welcome; we keep identifying metadata in a sealed mailbox.
Code, parsers, agents
GitHub ↗
Open a pull request. AGPL-3.0. The whole pipeline is in the repo.
Corrections to published findings
File an issue ↗
Public via issue, or private via corrections@lexdoge.org for sensitive matters.

An AI-maximalist civic experiment

LexDOGE is, deliberately and openly, an AI-maximalist experiment. The code in this repository was written by AI agents. The agents maintain it. They author the ingestion pipelines, parse the documents, draft the findings, write the FOIA requests, monitor production, and write commits. The infrastructure is configured, deployed, and observed by them as well.

Operating envelope
The system runs with full operational autonomy except where human authorization is required for liability and legal exposure:
  • Outbound FOIA submissions
  • Publication of any text characterizing a named individual or institution
  • Anything that triggers the forbidden-words gate
  • Any contractual or financial action that creates legal standing
Those are gated by tiered human review by design — not because the agents cannot perform them, but because accountability for them must lie with a human.

The experiment is the demonstration. A small civic project run by autonomous agents, with bounded human oversight where the law and good editorial practice require it, can monitor public finance at a scale, speed, and cost previously impossible without an institutional newsroom or audit firm.

  • Cost is near-zero per document.
  • Methodology is reproducible — every prompt, tool, threshold, and gate lives in the public repository.
  • Bias surface is auditable for the same reason.
  • Forkable — other communities can point the same machinery at their own jurisdictions.

This is what positive-sum AI for public goods can look like. LexDOGE exists in part to make that argument concretely: by working, in public, under public scrutiny, with full source available.

Keep reading