Publish the check, not the promise

How every number here
was produced.

Three sources, one join, and the exact commands to reproduce any figure on this site. If a check is not published, the claim it supports should not be believed.

The join

Three sources in, one dataset out.

Source	What it is	Rows	Authority
`source/portfolio.json`	The market view, as written by Kristen Hall	61	Authoritative for purpose, audience and revenue model
`data/apps.json`	The order of battle, carrying measured build state	72	Authoritative for build state, IP status and ceilings
`data/fleet.json`	Live HTTPS probe of every candidate address	438	Authoritative for what is reachable, at probe time only

The rule that governs every address on this site

A name that resolves proves nothing.

bornbetween2generals.com carries a wildcard DNS record. Every conceivable name resolves, including nonsense — so DNS cannot tell you whether a subdomain is real. An unconfigured subdomain has no certificate, so it fails the TLS handshake. That is the test used here, and it is the only one that means anything.

# does this subdomain actually exist? 000 means no certificate means no.
curl -s -o /dev/null -w '%{http_code}' -m 15 https://NAME.bornbetween2generals.com/

Being reachable at a matching name still does not prove a page is a given product. So each probe also captures the page's own <title>, and a product is only bound to an address when that title corroborates it. 25 portfolio rows live at a plausible address whose page calls itself something else; every one of them is shown as unconfirmed and none is counted as live.

Reproduce it

Run the checks yourself.

python3 scripts/probe.py known-subdomains.txt data/fleet.json   # registry C, from scratch
node   scripts/reconcile.mjs                                    # the join + the disagreements
python3 scripts/measure-hero.py                                 # every hero, WCAG on the 98th percentile
python3 scripts/gates/poison-hero.py                            # prove that gate can fail
node   scripts/gates/check.mjs                                  # every printed figure vs its source

The contrast gate above was itself wrong once. Under a heavier veil it scored a pure white hero at 6.2:1 and passed it — a gate that cannot fail is not a gate. It now fails white and near-white, and the poison test that proves it is part of the build. What it measures is the 98th-percentile brightness of the band the copy actually occupies, composited under the exact veil the site ships.

What this site does not do

Stated limits.

No figure here is a price, a quote or an offer. Revenue models describe intended structure only.
Every ceiling is BB2G's own arithmetic. None is third-party validated.
Build state is as measured on the order-of-battle generation date, not continuously.
The fleet probe is a point-in-time snapshot; an address that answered then may not answer now.
Nothing in the fleet can currently take a payment.
11 portfolio rows have no live address at all. That absence is stated, not hidden.