Steelmanning
Trains the discipline of engaging the strongest version of opposing views before responding.
Onramp · Foundation · Bridge Practice
The Codex Lens
Most disagreement is not actually with the other person's position. It is with a simplified, weakened version of it that is easier to reject.
This is the straw man, and it is the default mode of argument. It does not require malice. It does not require stupidity. It happens automatically, because when you are defending your own position, you have no incentive to make the opposing position look strong. Confirmation bias supplies a distorted version. Identity investment makes the distortion feel like accurate perception. The result is a person who sincerely believes they have engaged with the opposing argument when they have engaged only with their own projection of it.
Control feeds on this. In soldier mode, you construct the weakest possible version of every challenge, defeat it, and feel vindicated. Each victory reinforces certainty. Each reinforcement narrows the range of positions the mind can take seriously. Over time, you end up surrounded by straw men you built and beat, unable to see the real arguments you never actually confronted. This is how people become simultaneously certain and uninformed. They have won every argument in a room full of opponents they invented.
Decay feeds on it too, from a different angle. A person who has watched enough bad-faith argument, enough straw-manning from every direction, can conclude that understanding between positions is impossible. That disagreement is always performance. That nobody is actually trying to understand anyone else. This cynicism is the Decaying Mind's version of the same failure: it gives up on the project of genuine engagement because it has never seen it done well.
Steelmanning breaks both patterns. It is the discipline of constructing the strongest possible version of an opposing view before responding to it. You do not understand a position until you can articulate it so clearly and fairly that its proponent would say: "Yes, that is exactly what I mean, and those are my best reasons." Until you reach that point, you have not earned the right to disagree. You are arguing with your own imagination.
This is the only tool in the Onramp that bridges two Codex disciplines. As an individual practice, steelmanning is Foundation work: it forces you past your own confirmation bias and makes your own positions stronger by testing them against the best counterarguments, not the worst. As a relational practice, it is Bond work: it signals respect, creates the conditions for genuine dialogue, and makes it possible for the other person to actually hear your critique. A critique that begins with an accurate understanding of the other position lands differently than a critique that begins with a caricature.
The Concept
Steelmanning is the practice of articulating an opposing position in its strongest, most defensible form before critiquing it.
The contrast makes it concrete. The straw man takes the weakest element of the opposing view, exaggerates it, and attacks the exaggeration. The steelman finds the strongest element, states it accurately, and then responds to that. The straw man makes you feel like you won. The steelman tells you whether you actually did.
This is harder than it sounds, for a specific reason. To steelman a position, you have to temporarily set aside your own. You have to inhabit the other perspective well enough to find its best arguments, which means suppressing the reflex that treats those arguments as threats. You will resist this. Giving the best version of the opposing case can feel like conceding ground. It feels dangerous, because it might reveal that the opposing position is stronger than you thought. That possibility is exactly why the practice matters. If the opposing position is stronger than you thought, you need to know that. If it is not, you lose nothing by engaging its best version.
There is a distinction worth making between steelmanning and simply being polite. Politeness is a social performance. It asks you to nod respectfully at the other position and then continue holding your own. Steelmanning asks you to do real cognitive work: find the evidence that supports the other view, identify the values that make it coherent, locate the experiences that would make a reasonable person hold it. The output of politeness is a comfortable conversation. The output of steelmanning is an accurate map of the disagreement.
There is also a distinction between steelmanning and agreement. Steelmanning a position does not mean accepting it. It means understanding it well enough that your acceptance or rejection is based on the actual position rather than a distorted version of it. You can steelman a view and still disagree with it completely. In fact, the disagreement becomes more valuable when you do, because it is now a disagreement with the real thing.
One more thing. Steelmanning is not charity toward the other person. It is self-discipline. The primary beneficiary is your own thinking. A position that survives contact with the strongest version of its opposition is a position worth holding. A position that can only survive contact with straw men is a position you are protecting rather than believing.
The Practice
The diagnostic question is this: "Can I state the opposing position so accurately that its proponent would say, 'Yes, that is what I believe, and those are my reasons'?"
If you cannot, you do not yet understand what you are disagreeing with. Your critique, however sophisticated, is aimed at something that does not exist in the form you are attacking. Stop. Do more work. Ask the other person to correct your understanding until they confirm you have it right. Only then does your response count as engagement with the actual position.
Three practices build steelmanning from an occasional effort into a reliable discipline:
The articulation test. Before responding to a position you disagree with, write it out in your own words. Not a summary. Not a paraphrase that subtly weakens it. An articulation that captures the best evidence, the strongest reasoning, and the values that make it coherent. Then, if possible, show it to someone who holds that position. Ask: "Is this a fair representation?" If they say no, you are not done. If they say yes, you have earned the right to respond. In practice, you will often find that the act of articulating the position changes your understanding of the disagreement. The real point of contention is rarely where you thought it was.
The motivation question. For any position you find absurd, ask: "What would a thoughtful, informed person have to believe or value for this position to make sense?" This is the question that separates steelmanning from caricature. Almost every position that seems ridiculous from the outside makes sense from the inside, given certain premises. The premises may be wrong. But they are usually not stupid, and identifying them is the only path to a response that addresses the actual disagreement rather than your emotional reaction to it.
The reversal check. Apply the same standard to your own positions. If someone were steelmanning your view, would they get it right? What would they miss? Where would they make your argument stronger than you usually make it yourself? This practice reveals where your own position is under-articulated. If you have trouble imagining someone steelmanning you well, it may be because your own reasoning is not as clear as you think.
The failure mode worth watching for: performative steelmanning. This is when you recite the opposing position accurately and then dismiss it with the same arguments you would have used anyway. The articulation becomes a ritual that precedes the real response, which was never going to be influenced by it. The test is whether the steelman changes anything about your response. If it does not, if your critique would have been identical whether you steelmanned first or not, the practice is not operating. You are performing it without doing it.
In the Wild
A product designer was convinced that a competitor's approach to onboarding was wrong. In every design review, she critiqued it. A colleague challenged her: "Can you explain why they designed it that way, using their best reasoning, not yours?" She could not. She spent a day researching the competitor's design philosophy, reading their blog posts, watching their conference talks. She found that the onboarding was optimized for a different user segment than she had assumed, one with different technical literacy, different patience thresholds, different goals. Their design made sense for their users. Her critique had been aimed at a version of the product that did not exist. She still disagreed with aspects of the approach, but her critique sharpened. It addressed the actual tradeoffs rather than the imagined failures.
A couple had been arguing about finances for years. He wanted to save aggressively. She wanted to spend on experiences while they were young. Each thought the other was being irresponsible. A therapist asked each of them to state the other person's position as fairly as they could. He said: "You think we should enjoy life now because we cannot guarantee the future." She said: "You think security means not having to worry about money later." Both were surprised. Neither had heard their position described that accurately by the other person. The conversation that followed was different from every argument that had preceded it. They did not suddenly agree. But they stopped arguing past each other, because for the first time, each knew what the other was actually saying.
A policy analyst was asked to write a brief opposing a regulation he supported. His first draft was thin. He found the weakest arguments against the regulation and presented them as the best the opposition had. His supervisor sent it back: "This reads like you are trying to lose the debate on purpose. Find the strongest case against this regulation. If there is not one, we should be more confident. If there is one, we need to know what it is." His second draft was substantially better, and it changed his own view. The regulation still had merit, but it had a cost he had been dismissing because it was inconvenient for his position. The final recommendation included a mitigation for that cost. It passed review. The first version would not have.
Think of the position you disagree with most strongly right now. The one where you feel certain the other side is wrong. Now ask yourself: could you explain their best argument to them, in terms they would accept? Not their worst argument. Not the version that is easiest to reject. Their actual best case. If you cannot, your disagreement is with something you built, not something they said. The practice starts there. Pick the position. Do the work. Find out what they actually believe, and why a reasonable person would believe it. What you do with that understanding afterward is up to you. But whatever you do will be better for having started with the real thing.
Where This Comes From
The Codex did not invent steelmanning. It adopted and extended a practice with roots in philosophy, debate culture, and the rationalist community. What follows is the intellectual history and where to go for deeper study.
The principle behind steelmanning is ancient. Aristotle's practice of presenting and engaging the strongest version of opposing philosophical arguments before advancing his own set the template. Thomas Aquinas formalized this in the Summa Theologica (1265-1274), structuring each question by first presenting the best objections to his position, then responding to them. The Thomistic method is still the clearest historical example of institutionalized steelmanning: you demonstrate that you understand the opposition before you answer it. For anyone interested in the structural discipline of engaging counterarguments, the Summa remains instructive even for secular readers.
The modern term "steelmanning" emerged from the rationalist community, particularly through Less Wrong and related blogs in the early 2010s. It was framed as the opposite of the well-known "straw man" fallacy. Where the straw man weakens the opposing argument to make it easier to defeat, the steelman strengthens it to make your own position harder to break. The Center for Applied Rationality (CFAR) incorporated steelmanning into its workshop curriculum as a core rationality technique. Daniel Dennett articulated a similar practice in Intuition Pumps and Other Tools for Thinking (2013), drawing on Anatol Rapoport's rules for criticism: you must re-express the other person's position so clearly that they thank you, list points of agreement, mention anything you learned from them, and only then offer your rebuttal. Rapoport's rules and Dennett's version of them are the most accessible practical starting points.
John Stuart Mill made the case for steelmanning from a different angle in On Liberty (1859): "He who knows only his own side of the case knows little of that." Mill argued that even if a position is correct, the person holding it needs to understand the strongest objections to hold it as a living conviction rather than a dead dogma. This is the argument the Codex builds on. Steelmanning serves your own intellectual integrity before it serves anything else.
Two limitations are worth naming. Steelmanning can be weaponized. A person who steelmans an opposing position and then dismisses it with a wave can use the articulation as a display of superiority: "I understand you perfectly, and you are still wrong." This is the performative failure mode described in the Practice section, and it is worth watching for in yourself. The second limitation is scope. Steelmanning works well in intellectual disagreement where positions are articulable and evidence-based. It works less well in conflicts driven primarily by emotional hurt, power dynamics, or material interests where the real dispute is not about who has the better argument. Recognizing when a disagreement is about ideas and when it is about something else is itself a skill the Codex addresses through the Bond discipline.