Off-Path Forensic Research Loop
Find overlooked records, connect hidden relationships, then verify them in the field
- Difficulty
- Expert
- Time to result
- ~months to results
- Steps
- 7
- Confidence
- 97%
The Off-Path Forensic Research Loop seeks evidence in records that mainstream analysts rarely inspect. Investigators begin with a claim or accounting concern, identify the legal entities and jurisdictions involved, and retrieve lien records, securitization documents, and subsidiary financial statements. Those records are normalized and analyzed together so AI can compare accounting treatment, connect people across entities, and express patterns that specialists may struggle to communicate. The machine accelerates synthesis but cannot originate missing documents or guarantee correctness. Researchers must challenge sycophantic outputs, inspect original sources, and sometimes send people to retrieve records in person. The output is a verified contradiction or relationship suitable for deeper investigation.
Origin
Extracted from Masters in Business during Carson Block's description of Muddy Waters' document retrieval and AI-assisted accounting work.
Core principles
- 01Differentiated evidence often lives outside standard filings
- 02Local records can contradict consolidated financial claims
- 03AI accelerates synthesis but does not replace evidence gathering
- 04Human verification remains essential
- 05Entity relationships become more useful when examined across many records
How to run it
- 1
Define the Suspicion
Translate a vague concern into a claim that specific records could confirm or contradict, such as asset ownership, subsidiary performance, or lien release.
Pro tip Frame the question around a verifiable legal or accounting fact.
Watch out A broad request to find something dodgy encourages unsupported conclusions.
- 2
Map Entities and Jurisdictions
List subsidiaries, executives, lenders, counterparties, and locations connected to the issue.
Pro tip Assign consistent identifiers so similar names do not split one person or entity into multiple records.
Watch out Corporate name variations can create false separations or false matches.
- 3
Retrieve Overlooked Records
Collect UCC filings, securitization documents, overseas subsidiary statements, and other primary records outside the usual research path.
Pro tip Check what each jurisdiction requires companies to file publicly.
Watch out Do not assume the absence of a public document proves the underlying action never occurred.
- 4
Build a Searchable Corpus
Convert and organize the records so accounting terms, entities, dates, and transactions can be compared across sources.
Pro tip Retain links or references to every original document.
Watch out Poor document provenance makes later verification difficult.
- 5
Use AI to Connect and Express
Ask AI to identify recurring people, conflicting disclosures, and relevant accounting rules, then iteratively correct its interpretation.
Pro tip Use domain experts to tell the model when it has misunderstood an analyst's intended point.
Watch out Models may agree too readily or invent authoritative-sounding support.
- 6
Verify the Evidence
Read the decisive documents directly, confirm accounting citations, and retrieve unavailable records in person when necessary.
Pro tip Require independent support for any allegation important enough to publish.
Watch out AI output is not a substitute for original evidence.
- 7
Assess Publication Readiness
Determine whether the evidence is accurate, material, understandable, and strong enough to withstand legal and market scrutiny.
Pro tip Have legal and accounting specialists challenge the strongest claims.
Watch out Activist publication carries litigation risk even when the research is careful.
In the wild
Researchers reviewing asset-backed securities notice that public files do not contain documents showing that warehouse liens were released. Lawyers explain that private release letters are normal market practice. The team therefore reframes the issue: instead of alleging fraud from missing filings, it investigates whether the process permits forged releases or double pledging.
→ The inquiry moves from an unsupported accusation to a testable control weakness.
A US-listed company reports strong international operations. Investigators obtain locally filed financial statements from its overseas subsidiaries, organize them by entity and year, and use AI to compare the local numbers with consolidated claims. Human reviewers then inspect the original filings behind each discrepancy.
→ The team isolates verified contradictions that conventional SEC-only analysis missed.
Common mistakes
Uploading Only the 10-K
Standard filings contain the same information available to the rest of the market. Differentiated research often depends on records found elsewhere.
Treating Missing Paperwork as Fraud
A document may be private under accepted market practice. Confirm the legal process before interpreting a public filing gap.
Letting the Model Bully the Thesis
Sycophantic AI may endorse the researcher's suspicion too enthusiastically. Demand primary evidence and actively test alternative explanations.
Is it for you?
Best for
Experienced forensic researchers investigating complex companies, securitizations, and cross-border corporate structures.
Not ideal for
Casual investors without the time, legal tolerance, or access needed to verify primary evidence.
From the transcript
“one of the secrets. It's not a secret, but of what we do is we obtain documents from places people never look.”
“We love companies that have overseas subsidiaries because most overseas jurisdictions there are publicly filed financial statements.”
“But you still have to go out and do the legwork.”
From the episode
BONUS: Muddy Waters Capital Founder Carson Block
Carson Block