Ask San Diego’s public record a question. Get the answer, with the receipts.
Registrar, Census, FPPC, County, 18 city clerks and federal filings, pulled into one place so you stop hand-matching names across agency portals. Every answer carries the original source, the record ID and the date we retrieved it.
One place, not twelve portals. Hundreds of government sources in one record format, searchable together.
A receipt on every row. Source URL, page or row locator, checksum and retrieval date travel with each record.
Nothing hidden. What is loaded, what is still being collected and what cannot be told from the data are all shown.
—government sources inventoried
—source records held in the database
—verified exports open to assistants and download today
Checking the live catalog…
ASK. GET AN ANSWER. CHECK THE SOURCE.
This is what a cited answer looks like.
Four answers we checked by hand on November 3, 2026, one per beat. Pick one, or ask your own above.
YOU ASK
“Who is on the November 2026 ballot in Oceanside?”
Your assistant calls search_records on the candidate export.
It gets qualified candidates, offices and ballot designations.
Every row links to the Registrar’s own PDF and page.
THE ANSWER, WITH CITATIONS
The Registrar’s R740.03 candidate list (snapshot August 28, 2026) shows qualified candidates for Oceanside City Council districts 1 and 2 and Oceanside Unified trustee areas 1, 3 and 4. First ten rows:
Candidate
Office
Ballot designation
Michael Cline
City of Oceanside, Council District 1
Retired U.S. Army
Eric Joyce
City of Oceanside, Council District 1
Oceanside City Councilmember
Perry Alvarez
City of Oceanside, Council District 2
Automotive Repair Manager
Emily Gonzales
City of Oceanside, Council District 2
Director/Facilitator/Mother
Jerry Garcia Avila
Oceanside Unified, Trustee Area 1
Retired Lieutenant Commander
Nancy Licona
Oceanside Unified, Trustee Area 1
School Counselor
Joey Pearson
Oceanside Unified, Trustee Area 3
Professional Learning Director
Nicole Watson
Oceanside Unified, Trustee Area 3
Businesswoman/Mother
Raquel Riojas Alvarez
Oceanside Unified, Trustee Area 4
Governing Board Member
Jason Bethurum
Oceanside Unified, Trustee Area 4
Teacher/Father
More rows are available on the next page. Rows are candidacy occurrences as the Registrar published them, not resolved people; the same name in another dataset is not joined automatically.
“Compare rent burden in Encinitas and Vista with the county.”
Your assistant reads three Census tables for each place.
It applies the published formula for rent burden.
Each estimate keeps its 90% margin of error and survey window.
THE ANSWER, WITH CITATIONS
2020 to 2024 American Community Survey five-year estimates. Vista renters are more likely to be rent-burdened than Encinitas renters, while Encinitas has the higher median income and rent.
Place
Renters paying 30%+ of income
Median household income
Median gross rent
Encinitas
51.0%
$162,229 ± $10,306
$2,886 ± $115
Vista
59.6%
$94,975 ± $4,768
$2,249 ± $43
San Diego County
57.8%
$106,268 ± $949
$2,246 ± $11
Rent burden = (B25070 rows 007 to 010) ÷ (row 001 minus “not computed” row 011). Margins are 90% confidence. A difference between places is descriptive; it does not establish a cause.
Citation on every row
Source
U.S. Census Bureau, ACS 5-year table-based summary files: B25070 ↗, B19013, B25064
“What has SDG&E paid at the behest of state officials?”
Your assistant searches the FPPC Form 803 export.
It returns payor, official, payee, amount and date, as reported.
Each row points to the exact sheet and row in the state workbook.
THE ANSWER, WITH CITATIONS
Behested payments are gifts made at a public official’s request, reported to the Fair Political Practices Commission. First six rows for the SDG&E spellings:
Date
Official
Payee
Amount
2017-11-27
Ben Hueso Senate
California Latino Legislative Caucus Foundation
$15,000
2016-07-22
Lorena Gonzalez Assembly
Three Wise Men Veterans Foundation
$5,000
2016-09-29
Lorena Gonzalez Assembly
Jacobs and Cushman San Diego Food Bank
$5,000
2019-07-01
Connie Leyva Senate
Women in California Leadership
$15,000
2019-09-23
Connie Leyva Senate
Women in California Leadership
$5,000
2019-12-19
Shirley Weber Assembly
California Legislative Black Caucus Policy Institute
$20,000
More rows are available. The source spells the payor three ways, so a single spelling misses rows. Amounts are reported disclosures and are not summed into a total; a row is not an independently matched filing.
The 2025 point-in-time count for the San Diego City and County Continuum of Care found 4,191 people sheltered and 5,714 unsheltered, down from 4,495 and 6,110 in 2024. Of 4,633 people in the FY2022 permanent-exit cohort, 1,263 (27%) had a recorded return to homelessness within two years.
Measure
Value
Period
Publisher
Sheltered, point-in-time
4,191 people
2025
HUD, PIT by CoC (CA-601)
Unsheltered, point-in-time
5,714 people
2025
HUD, PIT by CoC (CA-601)
Year-round beds (ES, TH, SH)
5,010 beds
2025
HUD, Housing Inventory Count
Recorded returns within 2 years
1,263 of 4,633 (27%)
FY2022
Regional Task Force on Homelessness, SPM
City-identified homelessness spending
$218.2M
FY2020-21 to FY2022-23
California State Auditor, 2023-102.2, Table 2
A point-in-time count is one night, not annual demand. A recorded return is not proof that everyone else stayed housed. The spending figure is City-identified and was not independently validated by the auditor.
Sixty seconds, start to citation.Explainer video in production. Script and storyboard are in the repository.
HOW IT WORKS
From a question to a footnote in under a minute.
Type a question. Read the answer. Open the original document at the cited page. Copy a footnote-ready citation. Connect it to the assistant you already use when you want to go deeper.
Open Settings → Connectors, then Add custom connector.
Name it “San Diego Prosperity Observatory” and paste the URL above.
Leave authentication empty and save. Start a chat and ask a question like the ones above.
Claude Code (terminal)
Run the command below once. The server is read-only and needs no key.
claude mcp add --transport http prosperity https://san-diego-prosperity-observatory-sharper-ai.vercel.app/mcp
ChatGPT, Cursor and other MCP clients
Find the client’s MCP, connectors or tools setting.
Add a remote server with transport Streamable HTTP and the URL above.
Choose no authentication. The six tools appear as list_datasets, describe_dataset, search_records, get_record, get_coverage and research_question.
What we log: the questions and search terms sent to the tools, privately, to find missing sources. Not your conversation, account or IP address. Keep confidential sources out of your questions.
No assistant? Download the files.
Every public export is a verified JSONL file with a checksum, scope notes and source links. Open them in Python, R, a spreadsheet or DuckDB.
Counts come from the live source inventory and update as sources are added. A source is listed whether or not its records are loaded yet, so coverage stays visible.
Loading the source inventory…
—sources inventoried across the county, state and federal publishers
—sources with records loaded in the database
—verified exports open to assistants and download, with more in review
You can open the original. The source URL is the publisher’s own file, not a copy we edited.
You can find the exact line. The locator names the page, sheet, row or column the value came from.
You can prove nothing changed. The SHA-256 matches the file as retrieved on the date shown.
You can tell what the number is not. Units, universes and quality flags stay attached. A count of rows is never presented as a count of people or dollars.
You can cross-reference without being misled. Names are not merged across sources. A match is labeled a possible match until it is documented.
HOW WE WORK
Primary sources. Visible coverage. Reusable data.
Only primary sources. Government publishers and official filings. Nothing is typed in from a news story.
Gaps stay visible. A source we could not get is listed with the reason. An empty search is not proof that nothing happened.
Same method for everyone. Organizations and political actors are handled the same way, and we do not infer misconduct or causation.