Methodology
How We Curate
Shopify Dev Archive polls the Shopify developer ecosystem once a day, scores everything it finds with a language model, and sends out only what passes. Here's exactly how that works — and what happens to everything that doesn't make the cut.
Automated, not editorial
There's no editorial desk behind this. No human reads an item, decides it's interesting, and adds it by hand — every step from polling to scoring to publishing runs as code, applying the same fixed rules to every item, every day. That also means there's no slow news day lowering the bar, or a maintainer's personal taste raising it: the rubric below is all there is.
What gets polled
Four kinds of sources feed the pipeline, polled automatically:
RSS and Atom feeds
Shopify's own changelog and developer changelog, plus roughly a dozen agency, tooling, and community blogs and newsletters across the Shopify and headless-commerce ecosystem (including Liquid Weekly, This Week In React, and Baymard Institute).
GitHub Releases
Tag-by-tag tracking on Shopify's own cli, dawn, hydrogen, theme-tools, polaris, and liquid repositories, plus two closely related tooling projects, GraphQL Code Generator and React Router. A patch release only gets through if its notes mention something notable — a breaking change, a new command, a deprecation, a feature — so routine point releases stay out.
Hacker News
Stories about Shopify or AI with more than 50 points in the last 7 days.
A newsletter archive
I Only Speak Liquid's published back catalog.
Everything from every source is deduplicated and pooled together before scoring, so the same story linked from two places is only ever scored once.
How the LLM scores every item
Every new item — regardless of which of the four sources it came from — is scored 0 to 10 against one fixed rubric.
Scores well
- Directly relates to a tool, framework, or workflow used or taught for Shopify development
- Marks a meaningful shift in the Shopify development landscape
- Connects Shopify development to AI development workflows
- Is generating real discussion in the Shopify developer community
- Contradicts or validates something we've already covered
Filtered out
- A routine update with no meaningful impact
- Hype or speculation without substance
- Redundant with something covered recently
- Outside the scope of Shopify or AI-assisted development entirely
- Older than 7 days
- A GitHub patch release with nothing notable in its notes
- An AI story with no real connection to Shopify, e-commerce tooling, or agentic development workflows
Only items scoring 6 or higher pass. That threshold is fixed in code — nobody adjusts it by hand, on a slow day or a busy one.
Fixed sections and topics
Every item that passes lands in exactly one of three sections:
Platform APIs, Liquid, themes, and core commerce tooling.
CLIs, extensions, and the developer tooling ecosystem.
Where AI agents and commerce intersect.
and gets tagged with one to three topics from a fixed vocabulary of twelve — never something the model invents on the spot:
Filter aggressively, keep everything
Filtering only tells half the story. Nothing that's scored is thrown away: every item the pipeline evaluates, pass or fail, is permanently recorded in our database. But only the passing items — the ones scoring 6 or higher — become part of the public archive you can browse and search here, and only those go out in the digest. Items that don't pass stay in our internal records for our own analysis; they're not published anywhere on the site. So what you see here is deliberately, aggressively filtered — but nothing is silently deleted just because it didn't clear the bar.
See it for yourself — browse the full archive of everything that's passed the filter, or search it by meaning instead of keywords.