PLPublic LedgerControl plane

Public accountability, with memory.

Follow public money. Surface risk. Turn records into action.

Public Ledger helps watchdogs, journalists, attorneys, and civic investigators review public spending, connect evidence across sources, and convert suspicious patterns into concrete next steps.

24 signals surfaced5 FOIA packets ready4 governed skills live
1

entities monitored

Municipal pilots and future districts tracked in one accountability graph.

10

public sources linked

Official portals, packets, archives, budget files, and public evidence channels.

11

records in the corpus

Historical documents available for review, extraction, and pattern analysis.

24

saved accountability signals

Persisted review leads produced by the historical watcher.

5

FOIA packets ready

Request drafts prepared for user-led filing or later assisted delivery.

4

governed AI skills

Versioned skills with eval gates, promotion history, and rollback.

How it works

Built for real investigations, not just dashboards.

The alpha is designed to move from public record to actionable lead: ingest the source, review the history, surface a signal, prepare the request path, and keep humans in the approval loop.

Review history, not just headlines

Public Ledger starts with historical backfill so suspicious spending, recurring vendors, and policy drift can be traced across years instead of isolated meetings.

Turn leads into action

Signals become reusable FOIA packets, reviewer work queues, and accountable follow-up paths instead of dead-end summaries.

Cross-reference the full record

Budgets, minutes, packets, videos, portals, and filing instructions are handled as different evidence types and then tied together around entities, events, and relationships.

Keep AI under governance

Agents can improve, but only through versioned skills, eval runs, promotion gates, incidents, and rollback. No silent drift.

Evidence types

Each type of public data is treated the way it should be.

Public Ledger is being built for pattern recognition across different evidence vectors, which means tabular data, text, transcripts, and FOIA paths each get their own handling before they connect into a shared investigative model.

Tabular data

Budgets, contracts, ledgers, vote tables, and line items are normalized for joins, outlier detection, and exact comparisons.

Text records

Minutes, packets, staff memos, and reports are chunked, cited, and converted into extractable claims, entities, and events.

Video and audio

Meeting recordings become timestamped transcripts so what was said can be compared to minutes, later contracts, and follow-on actions.

FOIA and response paths

Request channels, filing constraints, templates, and returned records become a structured action layer, not just a reference note.

Pilot spotlight

Highland Park, Illinois is already seeded and under review.

The current pilot combines meeting records, budget and finance documents, public archives, and FOIA readiness so the app can test both historical accountability work and future monitoring.

What is live now

  • City of Highland Park is the first live pilot.
  • Historical review scanned 11 documents and surfaced 24 signals (14 high, 10 medium).
  • 5 FOIA request packets are already prepared for the pilot jurisdiction.
  • 3 of 3 managed providers are configured for the current alpha stack.

Recent signal examples

highHighland Park FY2008 budget final

Financial controls, reserves, or liability signal

Weighted watcher score 13 from 3 matched phrases related to fund balance, reserve policy, internal controls, accounts payable, or long-term liability language. Document type: financial_report.

highHighland Park FY2014 budget

Financial controls, reserves, or liability signal

Weighted watcher score 10 from 4 matched phrases related to fund balance, reserve policy, internal controls, accounts payable, or long-term liability language. Document type: financial_report.

highHighland Park comprehensive annual financial report FY2020

Financial controls, reserves, or liability signal

Weighted watcher score 7 from 2 matched phrases related to fund balance, reserve policy, internal controls, accounts payable, or long-term liability language. Document type: financial_report.

highHighland Park annual comprehensive financial report FY2022

Financial controls, reserves, or liability signal

Weighted watcher score 17 from 4 matched phrases related to fund balance, reserve policy, internal controls, accounts payable, or long-term liability language. Document type: financial_report.

Documents already in the pilot

Highland Park public permitting and inspections access noteother
Highland Park FOIA and public-record access notesother
Highland Park City Council recap - February 23 meetingmeeting minutes
Highland Park City Council source mapother

Governed AI

Self-improvement is live, but it is controlled.

Agents can refine skills from the beginning, but promotion is tied to versioning, evaluation, approval history, and rollback. That keeps the system learning without becoming opaque.

Current governance state

  • 4 governed skills are live in the registry.
  • 5 skill revisions are recorded with activation history.
  • The app is reading live Supabase state and Privy-backed identities.

What sign-up unlocks

  • Secure identity for reviewer, contributor, or operator access.
  • FOIA-ready workflows tied to your role and approvals.
  • Participation in the first pilot before broader public rollout.
  • Future access to governed AI outputs with human review in the loop.