A LEGO notebook · Oct 2026

The vocabulary

LEGO has made 9,461 different shapes. A 2,000-piece set uses about 230 of them. How does a designer choose?

Treat every shape as a word and every set as something written with them, and there's a dictionary to read: which words do the work, whether designers pick them on purpose, how each theme talks, and the years LEGO coined far too many. The main essay ran through this in two steps. This is the long version, with the controls left in.

Step 1 · The words

Fifty-five shapes make half of everything

Line up every shape by how many pieces of it LEGO has put in sets, and the curve falls off a cliff. The 1×2 plate alone is one piece in every 28. The top 55 shapes account for half of all pieces; the other 9,400 or so share the rest.

So a few words do most of the work. Is that design, or just what happens when some words are common?

Step 2 · The null model

Fit the designer yourself

Here's the test from the main essay, with the knobs exposed. Build a set by drawing pieces from LEGO's real catalogue, weighted by how common each shape is. That's the random baseline, and it needs far too many shapes. Now let a designer reuse: before each new piece, with some chance, take another of a shape already in the set.

Set how often the designer reuses in a 20-piece set, and how much more often in a set ten times bigger. The yellow line is real sets. Get your line onto it.

Mazzolini and colleagues showed in 2018 that uneven popularity explains a lot of how sets share parts, which is why the random line has to come first. Here it falls well short. Real sets look like a designer who reuses more and more as the model grows: walls of the same brick, roofs of the same slope.

That's the average designer. Each theme has its own accent.

Step 3 · Dialects

Every theme talks a little differently

Fit each theme on its own and two numbers come out. One is how fast its vocabulary grows as sets get bigger: double the pieces, and how many more shapes does it need? The fewer, the more it reuses. The other is how many shapes a typical 100-piece set uses. And each theme leans on words the others barely touch.

Words cost money, though. A new shape needs a new mould. A new colour is the cheapest new word, no mould at all, but it multiplies the stock of every shape made in it. In the late 1990s LEGO stopped counting.

Step 4 · The bubble and the extinction

Colours from 37 to 98, and back to 59

Between 1997 and 2004 the number of colours in use nearly tripled. Elements, a shape in a colour, more than doubled. About 4 in 10 of the elements that debuted in those years turned up in just one set before disappearing. Then came the turnaround: by 2007, 59 colours.

Here every colour is a row of studs in its own colour. Watch the band of new colours open in the late 1990s and close a few years later. And watch three greys and browns hand over in 2003 and 2004, the colour change builders still argue about.

Not every new colour died. Of the 58 on the chart that arrived between 1997 and 2004, 40 were still in sets after 2010. What collapsed was how many were in use at once, and how many elements LEGO coined to fill them. By David Robertson's account in Brick by Brick, the turnaround gave new shapes and colours a budget. Designers now spend yearly "frames," and each new shape or colour costs one.

The surprise is what happened next. Today LEGO uses two and a half times as many elements a year as at the bubble's peak, about 12,300 in 2025, spread across three times as many sets. Each set brings far fewer new ones: about nine new elements per set in 2004, about three in 2025.

So how does a designer pick 230 words out of 9,461? Mostly by reaching for the same few, more so the bigger the set. And since the extinction, by paying for anything new.

More LEGO notebooks

Sources and method

Rebrickable database, pulled Oct 24, 2026. Sets: everything except Duplo, Primo, Quatro, Gear, books, bulk bricks, service packs, education, collectible minifigures and Clikits; most charts use the 12,718 sets with 20 or more pieces. Pieces: each set's own inventory, without minifigure parts or stickers. A shape merges prints, patterns and mould revisions, using Rebrickable's part relationships and part numbers.
Null model after Mazzolini et al. (2018); the designer's reuse rule as in the main essay. Dialect fits: least squares on log distinct shapes against log pieces, per theme.
Robertson, Brick by Brick, on the 1997–2004 element count and the turnaround · New Elementary on frames
These counts come from a later pull and slightly different filters than the main essay's, so a few numbers differ a little.
LEGO® is a trademark of the LEGO Group, which does not sponsor, authorize or endorse this page.