Steelman Engine

What The Tool Does

The Steelman Engine performs a specific service: given any argument — one the user holds, one they oppose, one they need to defeat, or one they want to understand at its most serious — it produces the strongest defensible version. Most arguments are met in their weakest form, either because their defenders state them badly or because their attackers caricature them. The Steelman Engine short-circuits both failures by constructing the argument as its most capable proponent would.

The output is not an opinion. The engine has no view on whether the argument is correct. Its task is to produce the version worth engaging with — the version that, if it fails, genuinely fails, and that, if it succeeds, actually says something. From there the user can decide whether to adopt it, defeat it, or negotiate with it.

Where The Idea Comes From

The underlying intellectual move is old. John Stuart Mill in *On Liberty* (1859) insisted that anyone who knew only their own side of a case knew little of that side, and that a thinking person was obliged to state the opposing case "in its most plausible and persuasive form". The Russian-American scholar Anatol Rapoport, working on conflict theory in the 1950s and 1960s, codified this into what are now called Rapoport's Rules: before criticising a position, re-state it so clearly that the holder would say "I wish I'd thought of putting it that way" — and only then proceed to criticise.

The word "steelman" itself is more recent, emerging from the rationalist community around LessWrong in the 2010s as an inverse of "strawman". Julia Galef's *The Scout Mindset* (2021) is the current popular introduction to the discipline, but the practice is older than the term, and neither the word nor the rationalist milieu is required to do the work. The engine implements the principle, not the branding.

What It Can And Cannot Do

The engine can steelman positions across philosophy, politics, ethics, strategy, and policy. It can construct the strongest case for a view the user opposes, the strongest case against a view the user holds, and the strongest version of a view the user is trying to understand. It can handle positions the user finds repugnant — within limits. It is not a tool for manufacturing persuasive rhetoric in service of claims the engine judges genuinely dangerous, and it will flag cases where no steelman exists because the position rests on a factually false premise.

The engine cannot adjudicate. Two well-steelmanned opposing positions may both appear strong; deciding between them requires judgement, values, and further evidence the engine does not have. Its role is to raise the floor of the argument, not to declare a winner.