smart_money · security master, Stage A2026-09-01

CUSIP Succession, Working on Real Data

Five real corporate actions from the last two quarters, each independently verified against a news source, curated into the succession ledger, and checked against what our pipeline is showing right now versus what it would show with the fix applied. This is not a mockup — every number below is live production data.

5
events verified & curated
~$48.2B
currently misclassified as “new”
2,278
individual filer records affected
2
candidates correctly excluded

The five events

each confirmed against a company press release before being curated

Reorg
Brookfield Business Partners LP → Brookfield Business Corp — 1-for-1 unit-to-share conversion, 2026-03-27
272 filers · $11.4B
Redomicile
Indivior PLC (UK) → Indivior Pharmaceuticals, Inc. (DE) — scheme of arrangement, 2026-01-23
609 filers · $8.5B
Reverse split
Amcor PLC — 1-for-5 reverse split, 2026-01-15
1,327 filers · $28.2B
Merger survivor
RYVYL Inc. → RTB Digital, Inc. — merger closed 2026-05-12, ticker RVYL→RTB
30 filers · $11.2M
Name change
Mogo Inc. → Orion Digital Corp. — rebrand, no merger, 2025-12-29
40 filers · $1.8M

Exhibit A — identity continuity

Brookfield Business Partners → Brookfield Business Corp · JPMorgan Chase & Co, real position

What the pipeline shows today
G16234109 (old CUSIP)
  classified asEXIT
  shares → 01,013,814 → 0
113006100 (new CUSIP)
  classified asNEW
  shares70,300
Reads as: JPMorgan sold their entire Brookfield stake, then separately bought into an unrelated new security.
With the succession ledger applied
Canonical securityBrookfield Business (continuous)
Prior shares1,013,814
Current shares70,300
Classified asDECREASE
Reads as: JPMorgan trimmed a continuing position by ~93% — a real, large trade, but one position, not two fabricated ones.

JPMorgan's share count genuinely dropped well beyond the 1:1 conversion ratio — they were real sellers this quarter. The fix doesn't hide that. It just tells the true story: one continuing position that got smaller, not an impossible full-exit-then-brand-new-purchase in the same three months.

Exhibit B — identity continuity and the split ratio

Amcor PLC, 1-for-5 reverse split · BlackRock, Inc., real position

What the pipeline shows today
G0250X107 (old CUSIP)
  classified asEXIT
  shares → 0158,434,896 → 0
G0250X149 (new CUSIP)
  classified asNEW
  shares31,813,214
  value$1,264,575,260
Reads as: BlackRock exited entirely, then made a fresh $1.26B purchase.
With succession + ratio adjustment applied
Prior shares (raw)158,434,896
÷ split ratio (5:1)31,686,979
Current shares31,813,214
Difference+126,235 sh (0.4%)
Classified asUNCHANGED
Reads as: BlackRock's position was essentially flat through the split — the correct, unremarkable answer.

This is the sharpest version of the problem: without ratio adjustment, a stock owner who did nothing shows up as a $1.26B new buyer. This exhibit is also the one piece of the fix not fully wired yet — the ratio math above was done by hand from the curated event's ratio_old_to_new field to prove it out; the automated version is scoped for Stage B.

What the curation correctly refused to force

the review discipline is the point — a wrong merge would be worse than no merge

AstraZeneca — excluded, not a succession. The detector flagged AZN's ADR (CUSIP 046353108) against a new CUSIP (G0593M107) as a candidate pair. Research confirmed AstraZeneca completed a direct listing of ordinary shares on the NYSE in January 2026 — the new CUSIP is a second, concurrently-tradable security, not a replacement for the ADR. Collapsing these would have been a real error: two different instruments, sometimes different share ratios, both still live today.
Nine Energy Service — excluded, math didn't support it. The detector flagged a mass-exit/mass-new pair for this issuer. Checking the actual share counts: the “new” CUSIP shows more shares outstanding across holders than the “old” one — the wrong direction for a reverse split, and not explained by any confirmed corporate action. Held for further research rather than guessed.
SLMT / Brera Holdings — held, not force-paired. Covered in the earlier status report: three CUSIPs in overlapping, non-sequential use. The detector correctly reports all three as one review group instead of picking a plausible-looking but unverified “earliest → latest” pairing.

Where this leaves Phase 5

All five events are live in cusip_succession_events and cusip_canonical today. The before/after comparisons above were computed by a standalone script reading those tables — the live quarter-over-quarter pipeline that Qdrant, MCP, the mars mirror, and every current consumer read from is untouched. That's still Stage B: curate the rest of the backlog, wire the canonical lookup into the actual join, and automate the ratio adjustment shown by hand in Exhibit B.

What this demo answers is the open question from the last conversation: not “can we build this eventually,” but “does the fix actually work on real filings, including the messy ones.” JPMorgan, BlackRock, and the two exclusions above are the honest answer.