Anchored in the public record.
Meetings, agendas, staff reports, MOUs, term sheets, development agreements, permits, and related public records.
Meeting → filing → agreementThe record beforethe result.
Start with the meeting where the first terms appeared, the agreement that changed the deal, and the vote that moved it forward or stopped it. Every project record stays connected to its source documents.
Project file / illustrativeEvidence attached
01Municipal meetingsFragments located
02Agreements + termsTerms observed
03Entities + stakeholdersEntities normalized
04Project historyHistory connected
Claim → document → observed date134 supported fields
Processing scale
1,000-project reference- Meeting hours analyzed
- 4,000 hr
- Source-document pages analyzed
- 300,000
- Source data processed
- 27 TB
Reference cohort: 1,000 projects · 2,500 public meetings · 4,000 meeting hours · 300,000 source-document pages. Broader agenda and packet screening is additional.
Search finds what is published.Ultraground builds the record.
Public development information is fragmented by city, body, file type, meeting, and year. The terms that change a deal can sit in an attachment that search never indexes—or in a document never published online.
We source the missing record, process complete files and meetings, normalize the project, and connect each new source as it arrives.
Three connected layers
Each layer strengthens the next—and remains legible on its own.
One project, connected over time.
Sites, people, entities, terms, sentiment, decisions, and statuses normalized into a project-level view.
Entity → event → statusAnswers carry their trail.
Source references, record dates, observed terms, and freshness cues preserve the context behind each insight.
Claim → source → observed dateThe moat grows withevery project cycle.
As projects move, stall, restructure, or close, Ultraground preserves the sequence. That historical context improves the next comp, the next negotiation, and the next AI answer.