gossip · Data directory census
Every kind of data the house has touched, organised by the source-bucket taxonomy and tagged by how usable it is, feeding four instruments across past, present and future. The boundary is the load-bearing part: aggregate culture in, person-targeting refused.
The map, in one line. Kinds of data on the left, tagged by how much of each is actually live, pass through the harvest into the boundary gate, and only the aggregate remainder reaches the four instruments. Band X is what the gate turns away. Full row-level detail and the cultural-collections stack live in the house data directory.
Every source the house holds or has catalogued, one row each: its lane, its kind, its status, the type of open it is, and what may leave. The rows at egress never are counted and not listed; they name accounts and inboxes, and the boundary keeps them.
the register
·
| source | origin | lane | kind | status | open type | egress | items | last | used by |
|---|
Egress is computed from the lane’s open type, provenance and consent, never typed per row; the register that computes it is docs/data/source-register.json, built 2026-09-08. Items and last are the harvester’s counts for a registry row; a census row has none.
The map above is about what data exists. This is what obtaining it costs, measured once on 26 August 2026: one paged sweep of ten Mastodon tags, through four instances verified keyless, with every post the house declined to keep counted against the reason it was declined.
fig 2 . one sweep, ten tags, and what did not survive the refusals
1,026 kept + 787 refused = 1,813 seen, over 196 origin instances.
| tag | queried via | seen | kept | undiscoverable | bot | content warning | #nobot |
|---|---|---|---|---|---|---|---|
| #mechanicalkeyboards | fosstodon.org | 200 | 113 | 77 | 9 | 1 | 0 |
| #birding | mstdn.social | 200 | 113 | 87 | 0 | 0 | 0 |
| #sourdough | hachyderm.io | 200 | 128 | 67 | 0 | 1 | 4 |
| #fountainpens | mas.to | 200 | 156 | 40 | 0 | 4 | 0 |
| #vaporwave | fosstodon.org | 200 | 109 | 27 | 64 | 0 | 0 |
| #cottagecore | mstdn.social | 200 | 92 | 108 | 0 | 0 | 0 |
| #houseplants | hachyderm.io | 200 | 109 | 88 | 2 | 1 | 0 |
| #warhammer40k | mas.to | 200 | 88 | 105 | 2 | 5 | 0 |
| #speedrun | fosstodon.org | 200 | 111 | 40 | 49 | 0 | 0 |
| #gorpcore | mstdn.social | 13 | 7 | 6 | 0 | 0 | 0 |
| refusal | what it means | in this run |
|---|---|---|
| not-public | visibility is not 'public' | not counted |
| boost | a reblog is not the booster's writing | not counted |
| content-warning | sensitive set or spoiler_text present; the author covered it | 12 |
| bot | account.bot, self-declared, taken at its word | 126 |
| nobot | '#nobot' in the bio, a convention with no enforcement | 4 |
| undiscoverable | account.discoverable is false, the user's own opt-out | 645 |
generated 2026-08-26 · from the refusal-audit run · not hand-edited. Two declared refusals carry no count in this run.
The largest branch is a ruling, not a limit. The
discoverable field governs whether an account is listed in directories and
search; it does not govern reading. Treating it as a signal not to harvest is the house
being stricter than the platform asks. It is a ruling rather than a limit, which means
it can be changed by one.