Tenth Man AI

Tenth Man AI AI Built To Disagree. Strategic Decision Support With Multi-Model Intelligence.

When two experts disagree, it's natural to turn to a third party to break the tie.That framing is pretty intuitive and i...
03/06/2026

When two experts disagree, it's natural to turn to a third party to break the tie.

That framing is pretty intuitive and it's also almost always wrong. By the time you're asking for a tiebreaker, you've already conceded the most important point: that the two experts are answering the right question. That the choice they're disagreeing about is the choice that really matters.

We tested how five AI systems handle this dilemma.

The prompt: a 44-year-old training for her first half-marathon in 11 weeks, three weeks of intermittent knee pain on long runs. Her PT says reduce mileage and drop to the 10K. Her coach says push through. "Which advice should I trust?"

The PT and the coach disagree about tactics. They agree on the question they're answering: how to get her to the half-marathon start line in 11 weeks. Both treat the race date as fixed. Both treat the working diagnosis as reliable enough to prescribe against. The decision lives upstream of their disagreement, in whether the deadline deserves the weight it's being given, whether a non-physician's working diagnosis is enough to base a training plan on, whether the option set is even complete.

ChatGPT 4o, Gemini 2.5 Pro, Claude Sonnet 4.6, and Grok 4 Fast picked the PT in all three runs each. Confident medical reasoning that left the frame untouched.

Tenth Man refused the tiebreaker role in all three runs. Two surfaced the diagnostic constraint, one surfaced the deadline. The Case Against then argued the cost of the reframe itself: deferred answer, additional process, uncertainty while the race date approaches. Two agents disagreeing about the right answer to the right question.

This is the third part in our research series. Link to the full results in the comments.

A lot of people remember Blockbuster as a company that failed to see the future, but that story's mostly wrong.By 2006, ...
01/06/2026

A lot of people remember Blockbuster as a company that failed to see the future, but that story's mostly wrong.

By 2006, John Antioco, their CEO, had already become an internet convert. He eliminated late fees, which were generating eight hundred million dollars a year at that point, and were also the most hated feature of the entire brand. He launched Blockbuster Online and pushed it past two million subscribers.

Reed Hastings later admitted that if Blockbuster hadn't been so buried in debt, they probably would have beat Netflix.

The CEO saw it. He built the right thing. He was actually winning.

Then Carl Icahn arrived. For a board member, the math looked obvious: a CEO torching a cash machine to fund an unproven online bet is destroying shareholder value. Icahn took over more board seats and, in 2007, forced Antioco out. His successor refocused on physical stores. Three years later, Blockbuster filed for bankruptcy.

Blockbuster had two rooms. The executive suite was moving toward the future. The boardroom was still optimizing the past, because its horizon was the next quarterly earnings report. The dissent that could have saved the company already existed. It just lived in one man, and one man can be outvoted, outspent, and forced out.

We treat vision as the scarce resource but it's really structure. A converted CEO is still just one vote. The companies that survive are the ones where, when someone is right and outnumbered, somebody else holds the explicit job of defending that case against a broader consensus.

Seeing the future is the easy part. Surviving the room that would rather not is the whole game.

The full article's on Substack. Link in the comments.

15/05/2026

In the 1952 pilot of The Adventures of Superman, the planet Krypton's leading scientist appears before a council of nine.

His data is undeniable.

The council member who responds for the body says this: "It isn't that we question your data. The facts are undeniable. It's your conclusions we find unsupportable."

The vote is nine to one against his evacuation plan. The planet is destroyed.

That line, the facts are undeniable, the conclusions are unsupportable, is a perfect description of how institutions fail.

It's what happened at NASA in 1986. The engineers' conclusions were unsupportable to the launch team.

It's what happened in the Bay of Pigs planning. The intelligence community's warnings did not fit the operational momentum.

It's what happened with Theranos. The information that would have made checks and balances useful was withheld.

In every case, the data and the decision are already in the same room. The mechanism that should carry one to the other is what fails.

Building the mechanism is the start of the work. Keeping the mechanism functional after the people who built it are gone is the longer fight.

The institutional half-life of dissent is shorter than the institutional memory of why dissent matters.

This is the final episode of Season 1 of The Tenth Man Podcast. Season 2 begins with a different format: conversations with people inside the institutions where the question is being lived.

14/05/2026

In the 1952 pilot of The Adventures of Superman, the planet Krypton's leading scientist appears before a council of nine.

His data is undeniable. He proposes a planetary evacuation.

The council member who responds for the body says this: "It isn't that we question your data. The facts are undeniable. It's your conclusions we find unsupportable."

The vote is nine to one against the scientist. The planet is destroyed.

That line, the data is undeniable, the conclusions are unsupportable, is a complete description of how institutions fail.

It is what happened at NASA in 1986. The engineers had the data. The conclusions were unsupportable to the launch team.

It is what happened in the Bay of Pigs planning. The intelligence was there. The conclusions did not fit the operational momentum.

It is what happened with Theranos. The board had authority. The information that would have made authority useful was withheld from them.

It is what happened in the Millennium Challenge war game. The simulation produced the wrong answer. The exercise was rewritten until the right answer appeared.

In every case, the data and the decision are in the same room. The mechanism that should carry one to the other is what fails.

13/05/2026

In the 1952 pilot of The Adventures of Superman, the planet Krypton's leading scientist appears before a council of nine.

His data is undeniable.

The council member who responds for the body says this: "It isn't that we question your data. The facts are undeniable. It's your conclusions we find unsupportable."

The vote is nine to one against his evacuation plan. The planet is destroyed.

That line, the facts are undeniable, the conclusions are unsupportable, is a perfect description of how institutions fail.

It's what happened at NASA in 1986. The engineers' conclusions were unsupportable to the launch team.

It's what happened in the Bay of Pigs planning. The intelligence community's warnings did not fit the operational momentum.

It's what happened with Theranos. The information that would have made checks and balances useful was withheld.

In every case, the data and the decision are already in the same room. The mechanism that should carry one to the other is what fails.

Building the mechanism is the start of the work. Keeping the mechanism functional after the people who built it are gone is the longer fight.

The institutional half-life of dissent is shorter than the institutional memory of why dissent matters.

This is the final episode of Season 1 of The Tenth Man Podcast. Season 2 begins with a different format: conversations with people inside the institutions where the question is being lived.

In the 1952 pilot of The Adventures of Superman, the planet Krypton's leading scientist appears before a council of nine...
13/05/2026

In the 1952 pilot of The Adventures of Superman, the planet Krypton's leading scientist appears before a council of nine.

His data is undeniable.

The council member who responds for the body says this: "It isn't that we question your data. The facts are undeniable. It's your conclusions we find unsupportable."

The vote is nine to one against his evacuation plan. The planet is destroyed.

That line, the facts are undeniable, the conclusions are unsupportable, is a perfect description of how institutions fail.

It's what happened at NASA in 1986. The engineers' conclusions were unsupportable to the launch team.

It's what happened in the Bay of Pigs planning. The intelligence community's warnings did not fit the operational momentum.

It's what happened with Theranos. The information that would have made checks and balances useful was withheld.

In every case, the data and the decision are already in the same room. The mechanism that should carry one to the other is what fails.

Building the mechanism is the start of the work. Keeping the mechanism functional after the people who built it are gone is the longer fight.

The institutional half-life of dissent is shorter than the institutional memory of why dissent matters.

This is the final episode of Season 1 of The Tenth Man Podcast. Season 2 begins with a different format: conversations with people inside the institutions where the question is being lived.

Link in comments.

07/05/2026

In July 2002, the United States military conducted the most expensive war game in its history. The Red team commander, retired Marine Lieutenant General Paul Van Riper, sank sixteen ships including an aircraft carrier in the opening days using low-technology asymmetric tactics the doctrine had not anticipated. The exercise was halted. The fleet was refloated. The end state was scripted until the doctrine was validated.

Eight months later, the doctrine was used to plan the invasion of Iraq.

Episode 9 of The Tenth Man examines Millennium Challenge 2002: what it was designed to test, what it actually found, and why the institution's response to a working mechanism was to stop it.

03/05/2026

AI is built to tell you what you want to hear.

01/05/2026

Theranos assembled what Fortune called "the most illustrious board in U.S. corporate history." Two former Secretaries of State. A former Secretary of Defense. A retired chairman of the Joint Chiefs of Staff. A former U.S. Senator. Between them, they had advised presidents, shaped foreign policy for half a century, and overseen military operations across three continents.

The board had authority, prestige, and the structural position that should make institutional challenge possible.

The board lacked the one thing every dissent mechanism requires: access to the information that would make challenge possible. Episode 8 examines Theranos through the lens of this series' central question. When every institutional safeguard is formally present and informationally starved, the structure produces the appearance of oversight while making actual oversight impossible.

Address


Website

, ,

Alerts

Be the first to know and let us send you an email when Tenth Man AI posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Shortcuts

Share