The records tile says 448,724 while the page's own cards sum to 1,059,170

[realmarcin]

Reported by the site owner, who pointed at the strip figure on the live page.

The published page states 448,724 records. Its own ten Mech cards sum to 1,059,170.

card	records
TaxonMech	625,960
ProteinTraitsMech	408,978
CultureMech (published recipes)	10,657
HabitatMech	3,206
NaturalProductMech	3,115
MediaIngredientMech	2,951
AntibioticMech	2,939
TraitMech	630
CommunityMech	374
CellStructureMech	360
sum	1,059,170
strip says	448,724

The strip is short by 610,446, which is roughly TaxonMech's corpus. The tile was written when the fleet had nine members and TaxonMech's records were never added to it.

Three separate sets of numbers are in play and none agree:

• The strip's 448,724, typed into fleet_fragment.html.
• The card numbers, typed into mechs_template.md, refreshed in Refresh X-Mech overview from current published Mech sites #72.
• _fleet/data/prefix_census.json, which still holds the 2026-09-08 scan totalling 448,624 and has no TaxonMech entry at all. roots.py has no RECORD_GLOBS entry for TaxonMech either, so rerunning the census today would still not count it.

The strip's "43 vocabularies in dated census" has the same problem: a rescan of the current corpora gives 45, and that is before TaxonMech is included.

Two things are needed. TaxonMech has to enter roots.py so the scanning passes see it, and the strip has to be derived from the census rather than typed, so it cannot disagree with the cards again.

There is a definition question in the way, and it needs an answer before any number is published. The card figures and the census globs are counting different things for two Mechs:

Mech	card	RECORD_GLOBS scan today
ProteinTraitsMech	408,978	429,291
CultureMech	10,657 published recipes	6,286 merged recipes

Both are defensible corpora. Publishing one while the census computes the other is how the page got here.