How to Steelman an Argument

The internet runs on straw men. This is the opposite.

A straw man is the version of an argument you can win against—the dumbest reading of what someone said, propped up so you can knock it down. It's the engine of the whole gotcha economy: the ratio, the takedown, the “destroys” thumbnail. Winning by making the other side look stupider than they are.

A steelman is the exact reverse. You build the strongest version of the other side—on purpose, better than they built it themselves—and then you go up against that. If you're going to disagree, disagree with the best they've got. Here's the definition, if you want to steal it:

A steelman is the strongest, most honest version of an argument—often stronger than the people who hold it manage to make it. To steelman is to build that version first, on purpose, before you say a word against it.

Why almost nobody does it

Because the incentives are pointed the other way. Every attention machine we've built rewards the dunk: the clip that makes someone look foolish travels further than the paragraph that makes them look reasonable. Steelmanning is slow, quiet, and un-viral. It feels, at first, like helping the other team.

But there's a tell worth noticing. The people most confident that the other side is obviously wrong are usually the ones who've never once stated the other side in a form its holders would recognize. If you can't pass their ideological Turing test—if you can't argue their case well enough to fool them into thinking you're one of them—you don't actually know why you disagree yet. You just know that you do.

The Steelman Protocol

Steelmanning gets talked about as a virtue, which makes it easy to nod at and never do. So here it is as a procedure instead—five steps, in order. Run them like a checklist.

Step 1 · Find the version they'd sign

Restate the argument until the person who holds it would say: “Yes—that's exactly what I mean.” Not a version they'd quibble with. The version they'd sign their name to. If you can't get there, you're not ready to disagree; you're ready to keep listening.

Step 2 · Fix their mistakes for them

This is the step that separates a steelman from mere fairness. Repair the weak wording. Swap the bad example for a good one. Add the evidence they forgot to cite and the point they didn't think to make. You are, briefly, their best possible lawyer—working harder on their case than they did.

Step 3 · Say it out loud, un-hedged

State the steelman in your own voice, with none of the distancing language—no “they'd claim,” no “supposedly,” no smirk. If you can't say it like you mean it, you don't understand it yet. That discomfort is the whole exercise working.

Step 4 · Attack the strong version

Now—and only now—you get to disagree. Find the strongest objection to the strongest version. Aim at the steelman you just built, never at the straw you could've built instead. A win against the best version is worth something. A win against the worst version is worth nothing, and everybody watching can tell the difference.

Step 5 · Report the score honestly

Say what survived. Maybe your objection holds and the idea has a real hole in it. Maybe the steelman took your best shot and is still standing—in which case say that, out loud. Changing your mind isn't the failure state of this process. It's the point of running it.

What steelmanning costs you

I won't pretend this is free. Steelmanning is slower than dunking, and it will cost you things you might not want to pay:

• It looks, in the moment, like you're conceding. You're not—but a crowd trained on takedowns will read patience as weakness.
• It's genuinely dangerous to your certainty. Build the other side's case honestly and you may discover it's better than you thought. Sometimes you'll have to change your mind in public.
• It's more work. You have to understand two positions well enough to argue either, instead of one well enough to shout it.

Those costs are also the reason it's worth anything. A method that only ever flattered your existing opinion wouldn't be a method. It'd be a mirror.

A worked example

Watch the Protocol run on something people actually fight about—say, banning smartphones in schools. The point here isn't which side is right. It's that both sides deserve the strong version before anyone throws a punch.

Steelman for the ban: Attention is a skill, and it's being trained by the most sophisticated persuasion machinery ever built, all day, in every pocket. A classroom is one of the last environments designed to teach sustained focus; a device engineered to fracture it doesn't belong in the room any more than a slot machine would. Remove it and you're not being a Luddite—you're protecting the one thing school is for.

Steelman for keeping them: The phone is also how a generation reaches a parent in an emergency, coordinates a ride, manages a medical condition, and—done well—learns to self-regulate the exact temptation they'll carry for the rest of their lives. Banning the device doesn't teach that regulation; it just postpones the reckoning to a moment when nobody's supervising. You don't build discipline by removing every occasion to practice it.

Now you have two arguments worth having—instead of one strong one and one scarecrow. Whichever way you come down, you come down against something real. That's the difference the Protocol makes.

Steelman, charity, devil's advocate: what's the difference

Three ideas get used interchangeably. They're not the same.

principle of charity

Interpreting someone's words in their most reasonable light. A steelman goes further: charity reads the argument generously; a steelman actively improves it.

devil's advocate

Arguing a side you don't hold, often to provoke. A steelman is sincere—you're not performing opposition, you're building the real thing so it can be tested honestly.

straw man

The weak, distorted version you set up to knock down. The exact thing a steelman is the opposite of.

Who's telling you this

I'm Erik C. Olson. I've spent close to thirty years building software—often in high-stakes corners where a wrong assumption doesn't announce itself until it's expensive—and years in marketing, which is a paid education in how persuasion works from the inside. Both jobs teach the same lesson from opposite directions: the argument that feels most convincing and the argument that's actually right are two different things, and telling them apart is a skill. Steelmanning is the closest thing I've found to a reliable way to do it. It's the habit the whole show runs on. More about the show →

Try it on the next thing you're sure about

We build the strongest version of every argument—then pressure-test it. No straw men, no dunking. That's the whole show, and it starts in September 2026.

Want the one idea worth chewing on from each episode, written out? Get The Measure —subscribe here. It arrives when the conversation does, and the door swings both ways.