Product
Shopify sections: how we got the library to speak Liquid
"We used AI" is both true and a completely useless description of what actually happened. Here's the real version.
"We used AI" is both true and a completely useless description of what actually happened. Here's the real version.
Shopify sections are live, and new batches are landing every week.
We want to talk about how we did it, because the how is the genuinely interesting part — and because "we used AI" is simultaneously true and a completely useless description of what actually happened.
Worth being clear about what the job was.
Every one of these sections already existed. They'd been drawn, spaced, and argued over — sometimes years ago, always by a person. The hierarchy was settled. The type scale was settled. Nobody needed to design anything.
What we needed was a translation. Take a section built for Webflow and render it in Liquid, with Shopify's theme architecture, its schema, its own ideas about what belongs to the merchant and what belongs to the code.
Do that by hand, section by section, and it's a developer working full-time for a very long time. At the end you'd have a Shopify library frozen at whatever size the Webflow library was on the day you started — permanently behind your own product. We ran that calculation more than once and it never came out in favour.
Here's the thing that made it possible, and it isn't the AI.
Our sections are built to a consistent architecture. The same structural conventions, the same naming, the same relationship between a section and its variables — held across thousands of sections and several years. We were strict about that back when it was tedious, unglamorous, and completely invisible to anyone outside the studio.
Turns out that's the whole ballgame. You cannot automate the conversion of a mess. A library where every section was built its own way has nothing consistent to translate, and no amount of clever tooling fixes that. Every shortcut we didn't take three years ago is the reason this worked at all.
So when people ask what unlocked Shopify, the honest answer starts earlier than they expect.
Not "we asked an AI to convert our sections." We built a pipeline, and it took a while.
We trained it on our own structures — what our sections are, how they're assembled, what our conventions actually mean. We taught it what a good Liquid section looks like on the other side. Then we ran it, read the output, found where it was wrong, corrected it, and ran it again. And again. Until the conversion was reliable enough to trust at scale.
That's real work. It's just a different shape of work than it would have been a few years ago: less typing, more teaching and reviewing. The judgement didn't go anywhere — it moved.
The fair summary: the design is ours, the structure is ours, the training is ours. AI is the machine we built to do the repetitive middle, and it only works because of what we fed it.
We started by converting full collections rather than cherry-picking, which means it's a broad mix from the start — heroes, pricing, FAQs, feature sections, and plenty besides. The idea was to make it useful for building a whole store, not one part of one page.
There are several hundred already, and by the time you read this that number will be higher than anything we could sensibly put in a sentence. New batches land weekly. It's token-hungry, which is the honest reason it's paced rather than instant.
Same as everything else. Shopify sections sit on the same plans as the rest of the library: some are free, and Pro and Ultra get the whole thing. No separate tier, no add-on. If you're already a subscriber, they're already yours.
More every week, obviously. Beyond that, we want these in the MCP, so your agent can pull a Shopify section the same way it pulls a Webflow one. That's the version we're most excited about.
And if something's missing, tell us. That's genuinely how the Webflow library got to 2,500.
