This is Austin's public bus system — the real schedule, as its agency published it. We split that one collection into every format a business uses: tables, a spreadsheet, a database, a document, a JSON feed, a printed timetable. Download any face and inspect it with your own tools. Then ask the record a question — and watch the same answer appear in every face, because underneath them all there is one record.
The originals are byte-exact mirrors of the agency's published files (native). The other faces are derived from them 1:1 — recipe published, nothing added or altered (derived), so you can re-derive and verify any of them. Full hashes in SHA256SUMS.txt · recipes in RECIPES.md.
| Face | Format | What it is | Size | sha-256 |
|---|---|---|---|---|
| the original native | linked tables | CapMetro's full published feed — 70 routes, 24,830 trips, 2,346 stops (data.texas.gov) | 33 MB | 3890b0b38f1be233… |
| the 2018 edition native | linked tables | The same system as officially published in June 2018 — for time travel | 5.4 MB | 4210921a27072476… |
| stops.txt native | CSV | Every stop, extracted unmodified from the original | 400 KB | 1d8030ee7bb4610f… |
| network.sql derived | SQL | Routes, trips, and stops as a relational database — loads anywhere | 3.9 MB | a9bf78c7c199d4f6… |
| stops.xlsx derived | XLSX | The stops table as a spreadsheet — open it in Excel and count | 132 KB | fdfdb80f14ddc45c… |
| system.json derived | JSON | Agency, fares, and feed info re-encoded faithfully, counts included | 1.2 KB | 7986a1a02fe61e62… |
| route-801-timetable.pdf derived | Route 801's start times as a printed rider timetable | 3 KB | 22279267925e7799… | |
| fares-and-agency.docx derived | DOCX | Fares and agency info as a document | 1.2 KB | a4bc9766ff4ee283… |
verify any download: shasum -a 256 <file> (mac/linux) · Get-FileHash <file> (windows)
Click the button — the record computes the answer live. Then check every face you downloaded: the same number is already sitting in each one, because the format was never where the data lived.
The record holds both published editions of the system. Compare them, break them down — and ask for something the data doesn't contain, so you can watch it refuse rather than invent. It cannot make things up.
Most systems overwrite yesterday to store today. The record keeps every version it has ever been handed — whole, in place, and askable. No backup to restore, no export to dig out of an archive: pick a moment and ask it directly.
Between the agency's own 2018 and 2026 publications, total scheduled service nearly doubled. The record computes it from the two editions — press question 1, then question 2 above.
The rapid line that anchors the system runs about half the trips it did in 2018 — visible in the same two answers, and printed in the PDF face you can hold.
Press “what changed” and the record names every route added and removed between editions — plus 63 routes whose names drifted and an agency-ID scheme change nobody announced.
Stops moved, merged, and vanished across eight years. Every one is in the record, with the whole 2018 edition held intact beside today's — the two moments you can flip between above.
These buttons fetch a face back out of the live record — then your own browser hashes it locally and compares against the published fingerprint. The verification math never runs on our side.
Download any face, change anything — one digit in the spreadsheet, one time in the timetable — and drop your copy here. Your browser hashes it locally (it never leaves your machine) and compares it to the published fingerprint.
Also loaded and live right now: federal budget tables (XLSX), election filings (CSV · SQL), a government audit (PDF), an agency template (DOCX), and a safety regulator's feed (JSON) — all public, all pinned, all in the same record. Three things already found in them: health spending passed defense in 1994; the FY2025 audit reports $186 billion in improper payments across 15 agencies; and 1,914 candidates in the FEC files have no registered principal committee.