How accurate is this site?
This site is assembled by software (much of it AI-written) from the city's own portals. Software makes mistakes, so we measure ours in public: we drew 1,000 agenda items at random — with a fixed, published random seed, so anyone can re-draw the identical sample — fetched the city's live page for each, and compared field by field.
1,000/1,000city pages resolved
100.0%tracking codes correct
100.0%titles content-identical
48/48sampled attachments resolve
What we compared
- Tracking code (e.g. POR 2020 #173): 998 byte-exact of 1,000; 2 differ only in the city's own spacing (“ORD 2023 # 9”). 0 mismatches.
- Title: 184 byte-exact; 816 differ only by the city's trailing punctuation (their display appends a period; verified by inspection). 0 content mismatches.
- Meeting date: confirmed visible on 998 of 1,000 city pages. The 2 exceptions are itemized below — in both, this site dates the item to the meeting whose agenda carried it, while the portal shows the clerk's next-day action date after a past-midnight adjournment.
- Attachments (random subsample): 48/48 of the document links we mirror still resolve as valid PDFs at the city.
Every deviation found (4)
| Item | Type | Explanation |
|---|---|---|
| ORD 2023 # 9 | tracking format | the city displays a space after # (“ORD 2023 # 9”) our parser didn't expect — content identical |
| ORD 2025 # 8 | tracking format | the city displays a space after # (“ORD 2025 # 8”) our parser didn't expect — content identical |
| RES 2024 #82 | date display | we date it to its meeting (2024-04-29); the portal displays the clerk's action date, one day later (past-midnight adjournment) |
| RES 2024 #90 | date display | we date it to its meeting (2024-04-29); the portal displays the clerk's action date, one day later (past-midnight adjournment) |
Method, honestly
The comparison is against what the city's portal displays today — if the clerk edits a record after we capture it, that surfaces here as a deviation. The auditor is software too, so the code and raw results above exist for checking the checker. This page regenerates with every audit run. Found something wrong anyway? Every page has an error-report link in its footer — corrections are public.