Adversarial Source Corroboration
Compare opposed sources and trust only the overlap
- Difficulty
- Moderate
- Time to result
- ~ongoing to results
- Steps
- 5
- Confidence
- 96%
Adversarial Source Corroboration treats agreement between differently motivated sources as stronger evidence than repetition within one information camp. First, gather accounts from sources whose interests, audiences, or political positions conflict. Then trace each claim to determine whether apparently separate reports actually depend on one original source, a pattern Andrew Bustamante calls circular reporting. Promote only independently overlapping claims into the confirmed layer. Keep disputed claims in a living assessment rather than forcing a final conclusion. When the decision is not urgent, use time to collect more evidence and revise the assessment. The output is not certainty; it is a transparent map of confirmed overlap, unresolved disagreement, and the assumptions behind any provisional judgement.
Origin
Extracted from The Diary of a CEO
Core principles
- 01Opposed incentives make agreement more informative
- 02Repeated reporting may still trace back to one source
- 03Uncertainty should remain visible
- 04Time can improve an assessment when action is not urgent
How to run it
- 1
Map opposing sources
Select sources with materially different incentives, loyalties, or desired outcomes. Avoid treating several outlets from one information chain as independent.
Pro tip Include at least one source that would prefer each major competing conclusion.
Watch out Personal familiarity can increase trust without increasing objectivity.
- 2
Trace each claim
Follow repeated claims back to the earliest identifiable report or evidence. Label claims that all originate from one source as circular rather than corroborated.
Pro tip Build a simple claim-to-source table before judging credibility.
Watch out Volume is not independence.
- 3
Find independent overlap
Identify facts confirmed by sources with conflicting priorities. Give those points more weight while preserving differences in interpretation.
Pro tip Look for agreement on concrete events before motives or forecasts.
Watch out Agreement can still be wrong, so describe it as corroboration rather than proof.
- 4
Publish a living assessment
Separate confirmed facts, plausible interpretations, and unresolved claims. State what evidence would change the provisional judgement.
Pro tip Date the assessment so readers know it can change.
Watch out Do not turn missing overlap into a confident conclusion.
- 5
Use time deliberately
When action can wait, collect further observations and test whether the overlap persists. Update rather than defend the first assessment.
Pro tip Choose a review point before new evidence arrives.
Watch out Delay is a tool only when the cost of waiting is acceptable.
In the wild
An analyst sees the same casualty claim across many Western outlets. Instead of counting headlines, she traces them to one briefing, then checks whether Iranian state media independently reports the same deaths. Agreement between those opposed sources moves the event into the corroborated layer, while claims about motives remain provisional.
→ The assessment distinguishes a confirmed event from unverified explanations.
A founder sees many posts claiming a competitor is failing. He traces the posts to one anonymous thread, then compares customer complaints, the competitor's statements, and independent usage data. Only the decline in response times appears across sources with conflicting incentives, so he acts on that narrow finding rather than the broader collapse narrative.
→ The founder avoids a strategy based on circular reporting.
Common mistakes
Counting repetition as confirmation
Many reports may all repeat one original claim. Trace provenance before treating them as independent evidence.
Trusting only familiar people
A trusted relationship can make testimony feel credible without making it objective. Compare it with sources that have different incentives.
Forcing a final verdict
When evidence does not overlap, preserve uncertainty and maintain a living assessment instead of manufacturing certainty.
Is it for you?
Best for
It is best for consequential decisions made in a noisy or manipulated information environment.
Not ideal for
It is not ideal when an immediate safety decision leaves no time to gather additional sources.
From the transcript
“I want to take my information from as as far opposite sources as possible and then see where the information confirms itself, where it correlates.”
“It's one single source of information that comes out that gets multiplied over and over again.”
“So you can have a living assessment but you can't have a conclusion.”
From the episode
WW3 Threat Assessment: "Trump Bombing Iran Just Increased Nuclear War Threat" The Terrifying Reality