The Bear at Buffett's Own Meeting

In May 2013, Warren Buffett did something almost no executive does. He invited Doug Kass—a short-seller with an active bet against Berkshire Hathaway—on stage at Berkshire's own annual meeting, in front of tens of thousands of shareholders, to make the strongest case he could against the company. Buffett reportedly encouraged him to bring tough questions and, jokingly, to see if he could drive the stock down.

Buffett didn't do this for entertainment. He did it because an argument that has only ever been tested against weak opposition is an argument you don't actually know the strength of. Inviting your most credentialed critic to do their worst is the purest form of a discipline with a name:

Steelmanning. And before you file it under debate-club vocabulary: think about the last major proposal your team approved. Who was assigned to defeat it? If the answer is "nobody, but we discussed concerns," this post is about the gap between those two things.

What Steelmanning Actually Is

Steelmanning is the discipline of constructing the strongest possible version of an opposing argument before engaging with it. Not the version your opponent stated. The best version—the one they would have stated if they'd had more time to think, better evidence, and no rhetorical pressures. The version that, if true, would actually defeat your position.

The philosophical foundation is the principle of charity: interpret opposing arguments in their most favorable, most reasonable form. Assume the other person has good reasons for their position. Engage with what they mean rather than precisely what they said. Find the evidence they could have cited that they didn't.

The Canonical Protocol: Rapoport's Rules

Steelmanning has a canonical written protocol. The philosopher Daniel Dennett, in Intuition Pumps and Other Tools for Thinking (2013), codified four rules for criticizing an opponent's position, which he credits to the game theorist Anatol Rapoport:

Notice what the rules do: they make rebuttal a privilege you earn by demonstrating comprehension. The first rule is the steelman itself—if your opponent wouldn't sign off on your restatement of their case, you haven't understood it yet, and any attack you launch is aimed at a target of your own construction.

The Strawman Epidemic

In February 2013, Yahoo ended remote work for its employees. The memo—sent by HR head Jackie Reses and immediately leaked—argued that "speed and quality are often sacrificed when we work from home" and pointed to the value of hallway and cafeteria discussions. The public debate that followed was a masterclass in strawmanning on both sides.

Attacks on CEO Marissa Mayer ran along the lines of "she thinks people who work from home are lazy." Attacks on the critics ran along the lines of "they just want to avoid accountability." Neither is what either side actually argued.

The steelman for Yahoo's position: remote work severs the informal communication channels that enable rapid iteration, and Yahoo was a company in crisis that believed it needed serendipitous collision more than employee convenience. The steelman for the critics: if the turnaround strategy required innovation, eliminating flexibility would drive out exactly the high-performers with market options who chose Yahoo partly for it.

Both steelman cases are substantive and mutually illuminating. Neither dominated the debate. That is the strawman epidemic: in business as in politics, we systematically attack weakened versions of opposing positions because it's easier and feels like winning.

How the Pros Use It

Supreme Court Advocates: Moot Courts

Oral argument before the Supreme Court is the highest-stakes intellectual performance in American law, and advocates prepare for it with moot courts—practice sessions where colleagues, law professors, and former clerks play hostile justices and construct the most devastating available arguments against the advocate's position. The standard advice in advocacy training, including David C. Frederick's treatise Supreme Court and Appellate Advocacy, is to find the weaknesses in your own case before the justices do—because they will.

McKinsey: Red Teams and Blue Teams

McKinsey's published problem-solving guidance describes red-team/blue-team exercises: convening one group that is invested in the recommendation and another of impartial colleagues whose job is to pressure-test it before it reaches the client. The point is not politeness review—it is giving the strongest counterarguments a formal seat at the table while there is still time to act on them.

Amazon: Risk Sections and "Disagree and Commit"

Amazon's famous 6-pager memo format bakes the opposing case into the proposal itself: the narrative includes what could go wrong, which assumptions are load-bearing, and what the competition's response might be—so the counterarguments are on the table in writing before discussion begins. The surrounding culture supplies the other half: objections must be raised fully and explicitly before the decision, then everyone executes. We cover the full protocol in how McKinsey, Bridgewater, and Amazon turn disagreement into decisions.

Bridgewater: Seeking the Strongest Critics

Ray Dalio's core prescription in Principles is to find the most believable people who disagree with you and try to understand their reasoning—institutionalized at Bridgewater through recorded meetings and credibility-weighted debate. The firm's stress-testing culture had a famous outing: in 2008, while the S&P 500 fell roughly 37%, Bridgewater's Pure Alpha fund returned roughly +9%.

What the Research Actually Shows

Steelmanning's business case doesn't rest on anecdotes. Three findings matter:

But Doesn't the Research Also Say Devil's Advocacy Fails?

If you've read the group-dynamics literature, you have a legitimate objection ready: Charlan Nemeth's experiments found that an assigned devil's advocate—someone performing disagreement as a role—often doesn't improve thinking at all. Worse, it can produce cognitive bolstering: the group, having ritually survived the objection, becomes more confident in its original view, not less.

That finding is real, and it draws the line that matters. What fails is disagreement as theater. What works is authentic dissent—and steelmanning as defined here is not role-play, because step two requires actually going and finding the strongest evidence for the opposing view. You cannot fake research you are holding in your hand. This is also why Steve Jobs fired board members who never disagreed with him: he wasn't staffing a debate performance, he was buying genuine opposition.

The practical rule: never assign steelmanning as an attitude ("play devil's advocate for a minute"). Assign it as a deliverable ("bring the strongest documented case against this by Thursday"). The first produces bolstering. The second produces evidence.

The 3 Steelmanning Steps

Steelmanning is a learnable skill. Here's the framework:

Steelmanning in Product Decisions

Product decisions are where steelmanning pays fastest, because roadmaps are uniquely vulnerable to confirmation bias: teams build features that confirm existing beliefs about what users want rather than challenging them. Four exercises make the opposing case concrete before resources are committed:

How Argumentree Structures Steelmanning

The reason steelmanning is rare in organizations is structural, not motivational. Teams have no formal mechanism that requires building the opposing case. The default decision process—"propose, discuss, decide"—has no built-in steelmanning step.

Argumentree's argument tree structure changes this by design. Every decision tree captures pro arguments and con arguments with equal formality and rigor. The tree isn't complete until both sides are documented—not as a policy, but as a structural feature of the tool. And because the con side must be documented rather than performed, it sidesteps the role-play trap Nemeth identified: the tree demands evidence, not attitude.

Try It in Your Next Meeting

You don't need to overhaul your decision process to start. Try this in your next strategy meeting:

1. Before the meeting: Identify the main proposal. Assign one person to build the strongest possible case against it as a deliverable with evidence—not to play devil's advocate in the room.
2. In the meeting: Present the steelman case first. Ask the room: "Is this the strongest version of the objection? What have we missed?"
3. During discussion: When responding to the steelman, explicitly address its strongest points—don't pivot to the weaker ones that are easier to rebut.
4. In the decision record: Document the steelman and the responses alongside the decision. In six months, "why did we decide this?" has an answer that includes the strongest objections and why they were overruled.

Steelmanning is uncomfortable. It requires arguing against your own position with genuine effort, and it denies you the easy win over the weak version of the other side. But it's also the most reliable way to ensure your decisions can survive contact with reality—because they've already survived contact with the best possible case against them.

Buffett put a bear on his own stage. The question for your next big decision is simpler: who's yours?