Method

The apparatus of an argument is the visible object here, not a footnote. This page says what that means in practice and what it costs us to hold to it.

Every claim carries its own working

Sources, assumptions and confidence conventionally live in an endnote, set small, at the back. That placement is a decision about what the reader is trusted with, and we make the opposite one. A load-bearing claim travels with what it rests on, what it assumes, and what would falsify it. None of that is a footnote, a hover or a modal.

Only load-bearing claims get a block. If removing the sentence would not change the conclusion, it is prose. A brief where everything is a claim has no argument, only assertions.

The falsifier is never blank

Every claim states what would make it wrong, and whether we tested that. If we did not test it, the row says so. A falsifier we cannot name means we do not understand our own claim well enough to publish it, and the build refuses a brief that leaves one empty.

A conclusion cannot outrank its weakest input

Each brief ends with a load path: the conclusion, and the claims that carry it, drawn at the weight each claim can actually bear. The conclusion is rendered at the standing of the weakest claim in that path. The cap is computed rather than chosen, so it is not something we can talk ourselves past on a deadline.

Raising a conclusion means testing the weak claim or dropping it from the path. Dropping it usually means the conclusion can only describe rather than prescribe. Restyling the object is not one of the options.

Charts are real or they do not ship

There is genuine computation behind every mark. Where a density is shown it is a kernel density with a stated bandwidth, not a curve drawn to look like one. The dataset behind every figure is published alongside the brief, so anything we claim can be recomputed by the person reading it.

What we do not do

We do not use paid data. The sources we work from are public: SEC filings, FRED, Census, BLS, trade and patent data, regulatory dockets. That constrains which questions we can answer well, and we would rather be constrained than cite something you cannot check.

We do not promote a published claim quietly. A dated brief is a document of record. If a provisional claim is later tested, that is a new brief or a dated addendum, not an edit to the old one.