Under the hood of a broken conversation

Treat a stuck argument like a system that won't boot. You don't start by shouting at the screen; you start by reading the log. What were the inputs? Where did the two processes diverge? Which assumption is load-bearing, and which is just decoration?

Most conversations fail at the definition layer. Two people use the same word — "freedom," "fair," "natural" — to mean two incompatible things, and then spend an hour litigating the consequences of a disagreement they never actually surfaced. Find the word. Half the time, the bug is right there.

The next failure mode is the smuggled premise: a claim presented as a shared fact that one side never agreed to. Naming it doesn't end the debate, but it relocates it to where the real disagreement lives, which is the whole point.

Then there's the difference between a crash and a feature. Some disagreements are bugs — fixable with better information. Others are differences in values, and no amount of data resolves a values gap. Mislabeling one as the other is how arguments run forever.

The goal of the method isn't to win. It's to get to the smallest true sentence the two of you disagree about, and then look at it together. Sometimes you change your mind. More often you understand precisely why you don't. Both are upgrades over where you started.

None of this requires being right. It requires being willing to read the source before you ship the fix — which, it turns out, is the part most people skip.