Reading with the evidence in the margin

When a researcher reads a claim, the next question is almost always the same: says who? The honest version of writing answers that question in place. The evidence sits where the eye already is, not three scrolls down in a reference list that has lost all connection to the sentence that needed it.

This is the oldest idea in scholarly typography, and the web mostly abandoned it. Footnotes became endnotes. Endnotes became hyperlinks. Hyperlinks became tabs you open, skim, and never close. Each step moved the source a little further from the claim, until citing something and reading something became two different activities.[1]

The claim and its evidence should be neighbors

A sidenote is a footnote that refused to leave. Instead of sending you to the bottom of the page, it steps half a column to the right and waits there, aligned with the sentence that summoned it. You read the main thread without interruption; when you want the source, it is already in your peripheral vision.

This matters more for research than for ordinary prose. Research reading is non-linear. You move forward through an argument while constantly checking sideways: is this measured or assumed, primary or secondary, recent or inherited? The margin is exactly the right place for that sideways glance.[2]

Why the right margin, specifically

The left rail is already spoken for. A table of contents lives there because it answers where am I in the whole. The right margin answers a different question — what supports this part — so the two never compete for the same space.

Splitting the two roles across two margins keeps the center column narrow enough to read comfortably while still surfacing structure on one side and evidence on the other.[3]

It has to survive a phone

The honest objection to margin notes is mobile. There is no margin on a 390-pixel screen, so a layout that depends on one is broken for most of its readers.

The rule we follow: the margin is an enhancement, never a requirement. On a wide screen the note floats into the gutter. On a narrow screen the same note collapses inline, right after the paragraph that introduced it, as a small bordered aside. Nothing is hidden and nothing requires a tap — the reading order is identical, only the geometry changes.[4]

How this one is built

There is no runtime layout engine here. Each note is authored inline in the markdown source, right where it belongs, inside a ::: sidenote block. At build time the body is split into prose and notes; the prose flows down the center and each note floats into the right gutter with plain CSS.

Keeping it to one note per anchor point is what makes the pure-CSS version work — notes never need to dodge each other, so no JavaScript has to measure and reposition them. That keeps the whole page a static asset, which is the entire point of how this site ships.[5]

The result is a page that reads like a well-made handout: one clear line of argument down the middle, the apparatus of evidence kept close but out of the way. That is the reading experience research deserves, and the one most tools quietly gave up on.