The Tenth Man: A Useful Second Look
When everyone agrees, give one reviewer a useful task: find a credible alternative and a way to test it.
Body
Everyone agrees the system is ready. That is reassuring—until someone asks, ‘Ready for what, exactly?’ At MMG, the proposed Tenth Man role gives that question a place in the review.
The idea is simple: ask one person to examine the strongest plausible alternative to the shared explanation. The job is to improve the decision, not to win an argument. EMO does not currently have a demonstrated Tenth Man module; this is a method we want to test.
Make the claim small enough to check
‘The sound system works’ leaves plenty of room for misunderstanding. ‘The four planned monitor mixes have been tested at their intended outputs’ gives a reviewer something concrete to inspect.
The Campbell House production showed why that matters. The rental plan, tested console behavior and monitor fallback were different records. Combining them into one tidy story would lose useful information.
A second explanation, then a check
Imagine a silent microphone. The first explanation is a lost channel assignment. A reviewer asks whether the speaker is connected to the wrong output instead. Both ideas can be checked before changing the wider configuration.
Write down the decision, separate observations from assumptions, and ask for one credible alternative. Choose the smallest permitted check that distinguishes them. Record the result and any remaining uncertainty.
The reviewer supplies a perspective, not new authority. The responsible operator and existing permissions still govern the change.
Where the habit fits
MMG designs digital workflows as well as live production systems. In either setting, an agreed explanation can become a shortcut. A useful second look keeps the connection between the claim and the evidence visible.
EMO-64 asks how to describe a changing state. The Tenth Man asks whether that description stands up to scrutiny. These are complementary questions, not guarantees that more reviewers—or more AI agents—produce better decisions.
Test the review too
A trial should record helpful discoveries and the cost of finding them: time spent, false alarms and alternatives that led nowhere. Compare the method with an ordinary second-person review.
Success might mean finding a missing dependency, avoiding an unnecessary edit, or recognizing that more evidence is needed. Sometimes the most useful contribution is a well-supported reason to wait.