834 domains. Two reported. One real.
The Pulse universe more than doubles — 350 curated AI vendors grow to
834 domains with the addition of the Fortune 500. The crawl finds
two publishers. One is kineticgain.com, holding at 100/100. The other is a
Gatsby SPA catch-all serving its 404 page-context as application/json — a
false positive the current discriminator can't yet distinguish from a real document. The headline
number is two; the honest number is one, and the instrument note matters more than
the leaderboard.
The scaled universe
Issue #1 measured 37 domains. Issue #2 grew the universe to 350. Issue #4 carries that to
834 unique domains across 21 verticals — preserving the 350-domain AI-vendor cohort
and adding 484 Fortune 500 companies (industry-mapped where Wikipedia data allowed; otherwise
grouped under a Fortune 500 vertical with rank in the note column). The
universe.csv diff is reproducible end-to-end.
Fortune 500 vertical with
rank in the note. No buyer fabrication, no LLM-recalled brand lists.
What the crawl returned — and where it lied
corporate.charter.com (Charter Communications, Fortune 500 #71) reported a score of 82/100 — implying 9 of 11 Suite specs found a valid document.
It isn't a publisher. The site is built with Gatsby; its catch-all returns
a 404 page-context payload as application/json for any /.well-known/* path. The
response is structurally valid JSON, but its content is
{"componentChunkName":"component---src-pages-404-tsx",...} — the SPA's internal page-rendering
context. The probe discriminator at publish time accepted any 200 + valid JSON; it didn't yet require the
response to satisfy the specific Suite spec's shape (e.g., an aeo.json must declare an
aeo object, not a Gatsby componentChunkName).
The honest finding is still one real publisher (kineticgain.com).
The false positive is in the published data.json intentionally — Pulse readers should be
able to see the same evidence the instrument did, including its calibration gaps.
procurement-pulse-engine on main:
- Per-spec discriminator shipped (commit
5892d88). All eleven Suite paths now check a meaningful*_versionfield (e.g.agent_card_version,incident_card_version,decision_card_version). Re-probed live:kineticgain.comstill 100/100;corporate.charter.comdropped 82 → 0. The Gatsby SPA catchall no longer fools the instrument. All 15 engine tests pass. - kineticgain.com flipped to verified. All eleven dogfooded
/.well-known/docs are now ed25519-signed against a published public key atkineticgain.com/.well-known/pulse-signing.json. Engine probe now reports{ verified: 11, unsigned: 0, invalid: 0 }— the first verified row in any Pulse measurement, completing the Issue #1 load-bearing commitment about signature verification.
Issue #5 (the first true quarterly delta in August 2026) inherits both fixes by default. The numbers above remain as Issue #4 published them; this note is an additive editorial transparency layer, not a rewrite. The full locked-in calibration is published at /baseline/ as a between-issues note.
What hasn't changed
The headline pattern from Issues #1–#3 holds. Across every meaningful vertical — AI Platform, EdTech, HealthTech, FinTech, Enterprise SaaS, Data, Observability, Identity, Legal AI, CX/Support AI, Dev Tools, Marketing AI, HR AI, Retail, Government / Defense, Insurance, Automotive / Robotics, Energy — and the newly-added Fortune 500 vertical, the publication rate is still 0%. The mirror scaled; the field did not move.
Doubling the universe didn't surface a second real publisher. That is the most informative finding Issue #4 can deliver, because it sets up Issue #5 (the first true quarterly delta in August 2026): we now have a 2.4× larger surface against which to detect actual adoption when it appears.
What changes between now and Issue #5
- Per-spec discriminator in
procurement-pulse-engine. Each of the eleven specs contributes a minimal JSON-schema fragment (required top-level fields). A 200 with valid JSON only counts as published if the body matches. Catches Gatsby/Next/CRA catch-alls, generic 404-page JSON, and CDN error pages dressed as JSON. - Sign the dogfooded documents. The eleven
kineticgain.comfiles still readunsigned. Issue #5 ships ed25519 signatures + a public key atkineticgain.com/.well-known/pulse-signing.json, flipping the firstverifiedrow. - Push toward the long-arc 1,200 target. Issue #4 lands the Fortune 500 portion (~500 domains). The remaining 366 — top 100 K-12 EdTech, top 50 HealthTech AI, plus S&P 500 backfill — is editorial sourcing work, not engineering work.
- August 2026: the first quarterly delta. Three months of buyer pressure and vendor reaction compressed into one number. The instrument is now calibrated for a meaningful read.