Shopify is moving away from using a heavy JSON-based theme architecture toward a new format based on HTML and Liquid. This strategic shift is aimed at optimizing the performance of AI agents, allowing for a 93% reduction in code volume compared to the current Horizon system.
What Happened
Shopify is implementing a new syntax that includes composite block tags, React-like primitives, and enhanced Liquid logic with support for boolean expressions. The primary driver for this change is Shopify Sidekick—an AI assistant through which merchants have made 25 million edits this year. Moving from dense JSON structures to semantically rich formats significantly reduces the load on the LLM context window and decreases latency during data processing.
Context
Previously, Shopify's architecture relied on JSON-oriented structures designed for high data density. However, with the rise of autonomous agents, it became evident that such formats are less efficient for machine understanding and consume too many tokens, increasing inference costs and complexity.
Why It Matters for the Industry
There is a fundamental shift occurring in software interface design: instead of optimizing for human readability or data density, developers are beginning to design systems that are maximally convenient for LLMs (token-efficient and semantically rich formats like HTML/XML). This foreshadows the standardization of the 'LLM-readable API/UI' concept, where efficiency for machine understanding will become a key metric in architectural design.
Why It Matters for Users
For users and developers, this means a shift in the understanding of 'clean code.' In the near future, code quality will be determined not only by its convenience for humans but also by its token efficiency for AI. This will enable the creation of faster and cheaper-to-operate systems driven by autonomous agents.
Sources
Author
Look at AI, Editorial Staff