Recompute the tiers from the cleaned catalog, and drop FDX as first target

[kinlane]

Two changes: the catalog recompute from the standards work, and the recommendation change that follows from it.

From the catalog side

• 118 → 98 industry standards, derived rather than estimated. An admission test moved twenty entries out of the type — seven Java platform specifications, five horizontal formats, two things that are software rather than standards, and five buckets into a new umbrella type. Nothing deleted.
• Tiers are computed, not read by hand. Backfilled from a schema_forms field by fetching each body's actual artifact: A 30, B 36, C 13, D 12, seven unestablished.
• Tier D added — no wire format at all. Filing a methodology or an identifier scheme as Tier C implies a linter is failing to reach something that was never there.
• FIX and DICOM leave Tier C. Unlintable wire format and a machine-readable model published at source — a different problem, and a much more tractable one.
• Demand counts downgraded to unsourced, and AIS retracted: it read 45 companies, and sampling the matched text across 250 corpora returned "Air-Insulated Switchgear" every time.

The correction that follows

Three Tier A standards publish to members only — FDX, Berlin Group, Confirmation of Payee. Derivation works fine against a closed artifact; publishing the result does not, because a derived ruleset carries the thing it was derived from. They are the most lintable standards nobody can write a public ruleset for, and no amount of tooling fixes it — it is a licensing problem wearing a technical costume.

That removes FDX as the recommended first target, which both the research page and the post led with. CAMARA now stands alone: Tier A, open across 93 repositories, a live conformance question, and a Spectral ruleset for Open Gateway promised in December 2024 that still does not exist.

Adds "published openly" as a selection criterion — the criterion whose absence caused the original error.

Where the edits landed

File	Change
research/industry-rulesets.md	Recompute + dated correction on the first-target recommendation
_posts/…one-standard-twice…	Recompute summary + dated update note
_posts/…nobody-should-be-writing…	Worked examples move FDX → CAMARA; closed-artifact caveat added to bet one

All three are already published, so every change carries a dated correction note rather than being edited silently.

Jekyll builds clean; all six of today's posts still render.