The Crux: Finding What an Argument Cannot Survive Without
Most arguments are not lost on their strongest points. They are lost — or won — on one specific, quietly load-bearing premise that neither party noticed was doing all the work. In epistemics and rational discourse, this is called the crux: the one thing that, if you changed your view on it, would change your conclusion.
The concept was formalized in practice by the rationalist community at LessWrong, particularly through the work of the Center for Applied Rationality (CFAR) and the technique of Double Crux. Scott Alexander, Eliezer Yudkowsky, and others in the Slate Star Codex / LessWrong tradition used it extensively as a method for making disagreements legible and productive: instead of arguing everything at once, two people identify the one factual or empirical question they actually disagree on — the crux — and focus their attention there.
Why This Matters Beyond Rationalist Circles
The crux concept solves a structural problem in almost every public argument: most disagreements appear to be about one thing but are actually about something else. Two people arguing about immigration policy might appear to disagree about what's fair, but their actual crux might be an empirical question — does increased immigration suppress wages in the lowest income decile, or not? If they identified that, they might find they agree on values and disagree only on a factual prediction that, in principle, evidence could settle.
Without identifying the crux, debates tend to sprawl. Each side fires volleys at whichever point is easiest to attack, neither addressing the other's strongest argument. The conversation generates heat without progress. Finding the crux is a structural intervention: it turns an unfocused fight into a directed inquiry.
The Formal Structure
TellDear's Crux app formalizes this as follows. Given an argument, it:
- Identifies all premises the argument requires — including implicit ones that are never stated but must be true for the conclusion to follow.
- Assigns each premise a strength score: how widely accepted, how empirically grounded, how contestable it is.
- Identifies the crux as the weakest load-bearing premise — not the most important premise, but the most contestable one the argument cannot do without.
- Articulates what happens to the argument if the crux is false: does the conclusion fail entirely, or weaken, or require modification?
The key distinction is between weakest premise and most important premise. An argument might hinge on a premise that most people find obvious — that's a strong crux. The interesting cases are arguments that feel strong but rest on a premise that, once examined, turns out to be contested, uncertain, or simply assumed rather than established.
Double Crux and Productive Disagreement
The LessWrong community developed Double Crux as a collaborative technique: two people who disagree each identify their crux for their own position, then look for a shared crux — a single question whose answer would move both of them. When they find it, they've converted a debate into a joint investigation. They're no longer fighting; they're looking for evidence together.
This is radical in the context of most public discourse, where the goal is often to win rather than to update. The Double Crux method assumes both parties actually want to hold true beliefs more than they want to win. That's a significant assumption — but for the cases where it holds, it is probably the most efficient intellectual method available.
A good write-up of the technique is available on LessWrong: lesswrong.com/tag/double-crux.
What Crux Is Not
Finding the crux is not the same as steel-manning (building the strongest version of an argument), nor is it the same as looking for logical fallacies. The crux can be present even in arguments with no formal fallacies. It's also not about finding the most controversial premise — that would just be adversarial. It's about finding the premise whose truth value most determines the truth value of the conclusion.
It's also not a rebuttal. The Crux app does not tell you the argument is wrong. It tells you where the argument is most vulnerable — and therefore where honest inquiry should focus.
The TellDear Crux App
TellDear's implementation lets you paste any argument and get back a structured analysis: the crux highlighted, a one-line verdict, an explanation of why that premise carries the load, and what happens if it falls. The full premise list is shown with strength scores so you can see the whole dependency structure, not just the conclusion.
It's deliberately designed to be fast — one input, one output. The goal is to surface a structural insight you can act on in a real conversation, not to produce a research paper. Paste an argument you're in the middle of disagreeing with. Find out where it actually lives.