Alabama · Formation fee
$200
A value carries the date a person opened the source and confirmed it. That date belongs to the field, not to the page and not to the site.
last verified 2026-07-17
What happens when a number is missing?
This site says which kind of missing. A figure somebody opened a source and read prints with the date they read it. A figure nobody has opened yet says so, and a fact that was checked and found absent prints as a checked absence.
Those three are different answers, so they get three different cards. An unread value never enters a sort, never fills in as a zero, and never becomes a dash.
oldest record behind this page verified 2026-07-17
The three cards below are real fields from this build, found by one rule: the first match scanning states alphabetically, fields in the order a record sheet prints them.
California · DBA fee
unknown
Nobody has read this field at this pin. It cannot be ranked against a field somebody has read, so it stays out of every ordering instead of sorting as zero. See the full record sheet.
not read yet at this pin
Alabama · Annual report
none
Somebody checked and found there is nothing to pay. A verified absence is worth more than a number, and it enters arithmetic as a real zero.
verified absent 2026-07-17
Stable
A price is a contract
A filing fee or an expedite tier is something a state sells at a stated price. It renders as a value carrying the date it was read.
Volatile
A queue is an observation
How long a filing office is currently taking is weather, not a promise. It renders as a dated observation and never becomes an average.
no observation renders on this site yet: processing-time has no coverage row
How much is here, and how old is it?
Every figure comes from one source, the open us-llc-fees-dataset repository, pinned to a single commit for the whole build. The table below reads the record count that dataset states for each namespace, next to the number of records this build actually holds.
| Namespace | Records stated | Records in build |
|---|---|---|
| states.json | 50 | 50 |
| entitysearch-state-data/states/ | 51 | 51 |
| name-rules/states/ | 51 | 51 |
| dba-rules/states/ | 51 | 51 |
| anonymous_llc_available/states.json | not stated | 51 |
5 namespaces at pin @285f7ff
snapshot taken 2026-08-07
One row counts 50 rather than 51. The fee file predates the District of Columbia, whose figures this site resolves from the filing-office namespace instead, so no state is missing from a page.
How old are these figures?
Freshness belongs to a record, never to the site. Each state below counts once, under the oldest field this site prints for it, which is the same date its record sheet stamps.
July 2026 48 of 51
August 2026 3 of 51
Oldest record 2026-07-17, newest 2026-08-04.
What does the dataset say it is missing?
The source publishes its own audit rather than waiting to be asked, and recomputes these counts on every maintenance run. They are printed here as they stand at this pin.
-
0
statute citations still pointing at a commercial mirror instead of the state's own publisher. The mirrors that remain are the next card, and every one of them is deliberate.
-
20
mirror citations kept on purpose, in 4 jurisdictions that publish no free official copy of their code. Each one carries official corroboration and says so in its own notes.
-
20
of 51 states have an empty restricted-word list, and 17 of those are simply unread. This site prints no restricted-word list today, so the gap does not reach a page here.
-
0
cited links found unreachable when that was last measured, on 2026-08-06. Whether a URL answers is a fact about the network, so it carries the date it was checked rather than a running claim.
changes already known, not yet in effect
- Louisiana Articles of Organization and Annual report change on October 1, 2026 source
The figures on the page stay at today's values until that date passes. The amounts sit on the state's own row, so there is only one place for them to be wrong.
How does unfinished data stay off the page?
It is never copied into the build. The snapshot script reads the coverage table in the dataset README and takes only the namespaces listed there, so a directory that exists upstream but has no row is not fetched at all.
A folder full of records is not a publication signal. Somebody upstream has to say the namespace is ready, in the one place this build reads.
states.json
coverage row present
entitysearch-state-data
coverage row present
name-rules
coverage row present
dba-rules
coverage row present
anonymous_llc_available
coverage row present
processing-time
no coverage row
tax-rules
no coverage row
A namespace without a row holds back a record sheet field, and where the surface exists, a whole page. Both are already written and tested, and both render nothing rather than a smaller-looking number. The build stops loudly if a namespace opens before its field is implemented.
- /processing-time/ waiting on processing-time
- /franchise-tax/ waiting on tax-rules
Can a model read what a person reads?
Yes, because every figure is already in the HTML the server sends. Nothing on this site counts up from zero, and no table waits for a script before it holds numbers.
That is the whole trick, and it is also why the dates survive. An assistant quoting a fee from this page gets the fee, the date it was verified, and the office it came from.
/llms.txt
Site pointer for assistants
queued
/api/states.json
State index endpoint
queued
/api/llc/
Record sheet endpoints
queued
A machine-readable layer is not a substitute for the page. It is the same values in a second shape, and it arrives with the counters that say whether anything reads it.