LLOS.ai
Layout & Grids
🔹 Layout & Grids · Article 2 of 6

The Shared Edge

The straightest lines in a good layout are the ones that were never drawn. You just watched these ones find theirs.

One rail · fewer edgesEdges · not centresOptical · the eye wins

Every element on a screen extends invisible lines from its edges. When those lines agree — when a heading, a paragraph and a button all hang from the same left edge — the eye verifies order in one glance and gets on with reading. When they disagree, nothing looks broken, but everything feels faintly off.

That feeling is not taste; it is measurement. The eye detects a misaligned edge long before the mind can name it. Alignment is the craft of deciding which invisible lines your layout agrees on — and this article puts those lines in your hands.

01 · Why alignment

Lines you can’t see

Spacing, from the last article, sets how far apart things sit. Alignment decides something different: whether their edges land on the same lines. Two layouts can use identical spacing tokens and still read worlds apart, because one hangs together on shared edges and the other scatters.

The reason is mechanical. Reading is a chain of saccades — small jumps from landing point to landing point. Shared edges make every landing predictable, so the jumps cost nothing. Scattered edges make each jump a small search, and a hundred small searches read as clutter.

Hierarchy’s levers — size, contrast, isolation — decide what the eye visits first. Alignment is the skeleton underneath the levers: it holds the visits in order. Loud without aligned is chaos; aligned without loud is calm but flat. You need both.

Left: one rail — the eye’s returns are free. Right: four edges — every return is a small search. Multiply by every row on every screen.
The law

Shared edges make every landing predictable — the eye pays nothing to stay oriented.

02 · The shared left edge

One rail

In a left-to-right script the strongest line on any screen is the left rail — the single edge that headings, body text, media and actions all hang from. It is the cheapest order money can buy: no colour, no border, no extra pixel of ink. Just agreement.

The failure mode is never one bad edge; it is accumulation. A centred title here, a nudged image there, a button that drifted during a late fix — each defensible alone. Squint at the result and you can count the drift: five left edges doing the job of one.

Count them yourself. Same card, two modes — watch the rails light up.

Count the rails
Every distinct left edge lights up as a dashed rail. Fewer rails, calmer card.
left edges 1
03 · Edges, not centres

The return path

Centring feels generous — symmetrical, balanced, formal. For a short, symmetric moment it is exactly right: a headline, a tagline, a dialog title. The trap is applying it to running text, because centring aligns the one edge nobody reads from: the middle.

After every line break the eye swings back to start the next line. Give it a hard left edge and the swing is a reflex. Centre the text and every return lands somewhere new — the block grows a ragged edge on both sides, and the reflex becomes a hunt.

Trace the path yourself. The amber line is where your eye travels between lines.

The return path
The dashes trace where the eye lands at the start of each new line.
The law

Centre the moment; never the body.

04 · Optical alignment

The eye outranks the maths

Alignment sounds like arithmetic, and mostly it is — until a shape’s visual weight sits somewhere other than its bounding box says. A play triangle centred by its box reads as leaning left, because its mass crowds the flat side. A circle exactly as tall as a square reads smaller, because it fills less of its box.

The fix is the optical adjustment you met in the spacing article, now applied to edges: nudge until it looks right, then document the nudge. Type designers, logo designers and icon systems do this constantly — it is why professional icons come with their own overshoot rules.

Judge it yourself. Which mode looks centred — and which is?

Which one looks centred?
A triangle in a circle; a circle beside a square. Toggle between what the maths says and what the eye wants.
play button
same height · 54px
1
rail per screen, by default
0
pixels of ink alignment costs
4px
the nudge the eye just demanded
05 · Breaking it on purpose

The one broken edge

Once a layout holds one rail, breaking it becomes a lever. A pull-quote pushed past the text edge, a price hanging outside the card, an image bleeding to the viewport — against a disciplined rail, the one element off it earns instant attention. Misalignment is emphasis, but only where alignment is otherwise absolute.

You already felt this working: the amber pull-quote a few screens back hangs outside this page’s own rail, and your eye went straight to it. On a scattered layout the same quote would disappear into the noise. Deliberate, rare, documented — the same three words that governed spacing exceptions govern edges too.

One broken edge against a held rail — emphasis for free. Break it twice and it is just scatter again.
The law

A broken edge is loud exactly in proportion to how unbroken the rest are.

Seen in the wild

Same form, two skeletons

The exact same sign-up form — once on a single rail, once with every element choosing its own edge.

One rail
Labels, fields, button — one edge

Every element hangs from the same rail. The eye ticks down the form in a straight line; nothing needs re-checking.

Five rails
Every element, its own edge

Indented labels, drifted fields, a centred button. No single mistake — just five edges doing one edge’s job, and a form that feels harder than it is.

Where alignment meets the rest of layout

Spacing gave you the unit; alignment decides which edges those units measure from. Together they are the discipline half of layout. The next article adds the structure half: the grid — columns and gutters that turn one shared rail into a whole scaffold of them, built from the same spacing tokens.

The real references

Where the field agrees

Four sources worth reading in full — the systems and research this article stands on.

What people get wrong

Where alignment dies

Mathematically centred, optically offThe icon is dead-centre by the numbers and still looks low or left — because visual mass and bounding box are different things.FixTrust the eye over the ruler, nudge deliberately, and document the nudge so nobody “fixes” it back.
Margins drift per sectionEvery component owns its own container, so the hero starts at one x, the cards at another, the table nearly touches the viewport.FixDefine shared rulers — viewport edge, content edge, reading column — and make every section answer to them.
Baselines ignore each otherText in neighbouring cards sits at slightly different heights; each card typeset alone, none agreeing across the row.FixOne hang line for the row, one type beat underneath — cards are siblings, not strangers.
2–4px ‘close enough’Micro-misalignments too small to name accumulate until the page reads as subtly amateur — and nobody can say why.FixZero tolerance at the rails. Run the edge audit; merge or justify every edge. Small errors are the ones that compound.
The vocabulary

The Alignment glossary

Six words for thinking in edges, not instincts. Tap a card to flip it.

The nuance

Questions you actually ask

Train the eye

Run the numbers

Ten questions in mixed formats — multiple choice, multi-correct, fill-in, match, order, read-think-connect and write-your-own — each with layered hints: a nudge, the reasoning, then a deeper connection. Served from the question bank, never hardcoded.

Loading the question bank…
Question 1 of 10
Multiple choice
Keep going

Where this connects