Case study · not a scandal, a demonstration

Capital Graph

Behind Unusual Machines, Inc.'s board

Unusual Machines, Inc. (NYSE American: UMAC) went public 18 months ago and has since raised roughly $700M and attempted eight acquisitions. Its board looks clean on a résumé. Capital Graph queries The Ledger — the LLM's way into our fully linked graph — to show something a résumé doesn't: what happened to the other companies its directors have run.

ticker: UMAC sector: Robotics / Drone Technology IPO: Feb 2024 CIK 1956955
Before the case study

What is The Ledger?

What it is. Underneath Capital Graph is a full graph database — close to 200 tables of companies, people, investors, filings, and deals, sourced from SEC filings, regulatory disclosures, and market data. What makes it a graph rather than a pile of tables is that every one of them is linked to the others: a company to its directors, a director to every other board they've sat on, each of those companies to its bankruptcies, delistings, and M&A. The Ledger sits on top of that linked graph as the LLM's way in.

Why it's different. The Ledger does two things at once. It hands back pre-built answers already computed from those links — a director's full track record, outcome-classified, in one row. And every fact in it carries the real ID of the next hop, so the model can keep following the links as deep as the question actually needs. Most data offers one or the other: a summary with no way to go further, or raw records with no synthesis. The links are what let The Ledger be both.

Why it matters for an LLM. A model working from web search has to guess how many searches are enough, and can't safely connect what it finds. Working from The Ledger, it doesn't start from scratch and it doesn't stop at a static answer — it reads the pre-built summary, then keeps digging along the same connected data, hop by hop, with every step verifiable. The case study below is exactly that, run live.

Step zero — the baseline anyone can get

What a plain lookup shows

This is roughly everything a company profile, a stock screener, or one good web search returns in a few seconds. It's accurate. It's also the entire ceiling of what a single-shot lookup gives you.

Company
Unusual Machines, Inc.
HQ
Orlando, FL
Industry
Drone Technology
CEO / Chair
Allan Evans
Directors on file
7
Capital raised
$719M
Layer 1 — the company's own history

An aggressive roll-up, assembled in one call

Eight acquisition events chained off one company_id — dates, dollar amounts, and deal status pulled from real 8-Ks and press coverage, not one article at a time.

May 2026
Acquiring Upgrade Energy pending
$52.0M — battery/power systems for drones, adds domestic manufacturing capability
Jun 2025 – Sep 2025
Acquired Rotor Lab Pty Ltd completed
$7.0M — Australian motor engineering, cross-border deal
Jun 2025
Attempted acquisition of Aloft, Inc. terminated
$14.5M announced, deal did not close
Dec 2023 – 2024
Spun out of, and financed by, Red Cat Holdings (RCAT)
$20M + $3M related transactions; Red Cat is a named investor in UMAC's own $4M debt round, Jul 2024
Feb 2024
Acquired Fat Shark Holdings + Rotor Riot LLC completed
FPV goggles + drone-parts e-commerce — the deal that built UMAC's current product line
Layer 2 — the part a company profile never shows

Who runs it, and what happened at the other companies they ran

Every director's name, chained through donor_person_track_record, which returns their entire board/officer history elsewhere — each company tagged with a real, independently-sourced outcome. This is the layer no bio page or LinkedIn profile assembles for you.

Allan Evans
CEO & Chairman, UMAC
person_id 69c68763-bd19…
delisting-distress Previously CEO of Red Cat Holdings, Inc. (RCAT) — Red Cat received an NYSE American deficiency notice in May 2022 while he held that role. Source: SEC 8-K item 5.02 + delisting_events_v2, methodology bios-8k-502-v1
Sanford Rich
Director, UMAC · CIK 1362039
person_id da623af1-02f1…
credential Sitting member of the PCAOB Investor Advisory Group (2022–2026) — a real regulatory-body appointment, not a company relationship. Source: SEC-extracted board membership, filing 0001683168-26-006282
acquired Officer/director at InsPro Technologies Corp, which was acquired in Dec 2019.
delisting-distress Officer/director at Aspen Group, Inc. (ASPU), which received an NYSE American deficiency notice in May 2022.
Jeffrey Thompson
Director, UMAC (since Dec 2025)
person_id a5254c4d-28c4…
unverified lead A DEF 14A "other directorship" disclosure names him as a board member of Panacea Life Sciences Holdings, Inc. (PLSH) — flagged at 0.5 confidence as a candidate relationship, not yet corroborated by a second source. This is the "keys to go deeper" half of the row — a real lead to run down, presented as exactly what it is: unverified.
Robert Lowry & Brandon Torres Declet
Directors, UMAC
No other board, officer, or founder relationships on record anywhere in The Ledger — a genuine clean read. Capital Graph shows it here on purpose, rather than manufacturing a story where there isn't one.
The mechanism

How Capital Graph built this, in four calls

Nothing above required a human to know Sanford Rich's name in advance, or to guess that Red Cat Holdings mattered. Capital Graph walked The Ledger to find the people, then walked it again to find their history.

1
Resolve the company.
"Unusual Machines" → company_id 2478d37f…, CIK 1956955. One call, one strong ID to chain everything else off.
2
Walk the governance edges.
Every board_member_of / officer_of edge landing on that company_id — seven real, currently-active directors, no name typed in by a human.
3
Pull each director's full track record.
donor_person_track_record(person_id, include_outcomes=true) — one call per person, joined live against bankruptcy, delisting, and M&A outcome tables. No director required a separate investigation.
4
Hand back the keys, not just the answer.
Every row still carries its own company_id / person_id — Panacea Life Sciences, Red Cat Holdings, InsPro Technologies are all one more call away, at whatever depth is actually needed.
Why an LLM alone doesn't get here

The comparison that matters

Nothing above is secret. Every fact is in a public SEC filing somewhere. The gap isn't access to information — it's assembly.

A web search / plain LLM

  • ✗Finds UMAC's own news coverage fine — the acquisitions are real press events, individually searchable.
  • ✗Has no reason to search "Sanford Rich Aspen Group delisting" unless it already suspects something — there's no trigger.
  • ✗Would need 7 separate, correctly-targeted searches (one per director) just to check for outside history — and a person search often returns the wrong same-named person.
  • ✗Has no way to surface the Panacea Life Sciences DEF 14A disclosure — that fact lives inside a dense regulatory filing, not a news article.
  • ✗Produces a narrative summary with no chainable IDs — nothing to click into and verify or extend.

Capital Graph

  • ✓Starts from one identifier and mechanically walks every governance edge — no director gets skipped because nobody thought to search for them.
  • ✓Every outside relationship is independently outcome-classified against real bankruptcy/delisting/M&A tables, not inferred from prose.
  • ✓Confirmed facts and unverified leads are tagged separately — the Panacea lead is shown AS a lead, not smuggled in as fact.
  • ✓Same-named-person collisions are resolved by CIK/employer, not left to a search engine's ranking.
  • ✓Every row keeps its own ID — a reader (or another agent) can go one hop further on any thread, instantly.

"A search result is either a link — no synthesis — or a summary — no links back out. This row is both at once: a dense, pre-synthesized answer that also hands back the keys to go deeper."

— the core thesis this case study exists to prove, in a company's own words

Nothing in this report was cherry-picked to manufacture a scandal — two of UMAC's seven directors came back completely clean, and that's shown honestly rather than hidden. The value isn't "gotcha." It's that Capital Graph produces this level of diligence, on demand, for any company someone asks about — not just the ones we happen to already suspect.