Pipeline status

This page is written by the pipeline itself on every run. It is the evidence that the site maintains itself.

One row per run. A day can appear more than once: re-running a date is safe and writes nothing new, which is what a repeated day with zero new events means.

Data current through

2026-09-11

Currency

current

Last successful scan

2026-09-12 07:53 UTC

Overall

OK

This dataset is current. The newest filing day it holds is 2026-09-11, which is current. The pipeline fails closed: a run that trips an integrity check commits nothing, so what you see is the last state that passed every check rather than a partial or doubtful one. The workflow history shows whether a run has been failing.

Run it nowneeds your GitHub sign-in

Open the link, then use Run workflow on the right and leave the box blank — blank means “catch up on everything not yet scanned”, which is almost always what you want. A scan takes a few minutes and the site republishes itself when it finishes. If something looks wrong rather than merely late, the runbook covers what each kind of failure means and what to do about it.

Running in deterministic mode — no AI commentary. Signals come from SEC item codes and filing comparison only, at zero cost. Adding an ANTHROPIC_API_KEY secret enables plain-English analysis with no code change.

Self-checks 21 ok

These run automatically at the end of every pipeline run and assert the properties this site depends on. They are published here rather than logged, because nothing else is watching.

• Pipeline is current last complete filing day 2026-09-11, which is everything EDGAR has published ok
• Recent runs completed 10 runs recorded, 40,554 index rows read ok
• SEC access no refusals in the last 10 runs ok
• Every entry links to its own filing 4085/4085 link to the accession they report ok
• Every entry states how it was found 4085/4085 name a detection method the pipeline implements ok
• Item codes match the signal they produced 480 entries from SEC item codes, 0 carrying the wrong code for their signal ok
• Comparisons are like-for-like 964/964 compare the same form type against an earlier filing ok
• No duplicated entries 4085 unique entries across 4085 rows ok
• Quotes open with the substance 0 of 2592 begin mid-sentence, 0 do not open at the substance ok
• Quotes end where the filing does 0 of 2592 end mid-sentence with no ellipsis to mark the cut ok
• No entry contradicts its own quote 0 of 904 going-concern and material-weakness entries quote a passage that negates the state they assert ok
• Auditor names are firm-shaped 420 firm names across 87 firms; 14 spelled more than one way by filers ok
• Home page window showing the most recent 150 of 4085, as stated on the page; the full record is on the signals pages ok
• Days scanned in full 60 recent day(s) compared every periodic report filed ok
• Company size index 5,158 companies, refreshed 0 day(s) ago ok
• Follow-on rates computed 2361 companies, 17,970 historical events ok
• Page weight heaviest page companies.html is 85 KB over the wire (limit 250 KB), 2378 pages checked ok
• Firm labels are consistent 87 firms on the auditors page, one label each ok
• Filter options match the page 740 pages with a period or form filter, 0 option(s) that filter to nothing ok
• Data file weight heaviest file events.json is 829 KB over the wire (limit 4096 KB), 6 file(s) checked ok
• Pages agree all 2378 pages built 2026-09-14 08:28 UTC, data through 2026-09-11, 4,085 events ok

Last checked 2026-09-12T12:56:37+00:00.

Recent runs

Newest first by when the run executed, not by filing day. Each night’s scan processes the latest filings and then fills a chunk of older history, so the two kinds interleave.