AI NewsProductivityAnnouncement
Shopify writes 11x more mobile code with an AI attack-surface map
Shopify's Michael Hughes writes that mobile code volume rose 11x year on year. The team moved its security review to run on every patch.

Why it mattersA team producing more code per week needs a review that scales with it. Shopify shows one working design.
A security team that reads a big pull request slowly is the step that stops a team from moving fast. Shopify's mobile team now produces 11x more code per week than it did a year ago. The old security review did not fit that cadence. Michael Hughes writes on the Shopify engineering blog that the team changed its review. It now runs on every surface, every patch. Across a recent 12-week rewrite of the Shopify app, the team averaged about 434,000 lines of code changed per week. The year before, the number was 39,000. Both figures are Shopify's own.
The piece sets out four practices. They sit on top of standard mobile app security guides from OWASP. None of them are new tools. They are a new review shape.
A written map of the attack surface
Shopify writes down everything security-sensitive in the app. The list is grouped into surfaces. The surfaces named in the post are deep links, WebViews, stored data, permissions and dependencies. Each item on the list carries one of five verdicts: Fix (with a severity), Tradeoff, Secure, Fixed, or Debug only. A reviewer can see at a glance which parts of the app are known-good. They can also see which parts need a decision.
Reusable skills across humans and agents
Security best practice is written down as skills. There is one skill per surface. A skill describes what to inventory, what good practice looks like, and what needs checking. Skills live outside any single tool. They run inside a scheduled review, a one-off check, or an agent workflow. The common base lets the team reuse the same rules across humans and agents.
Review at the route level
Shopify now breaks review into small units. A reviewer looks at one route, every parameter it reads, and the code that reads each parameter. The small unit drops a reviewer's working context to something a model can hold at once. It also matches how the agents are changing code, one slice at a time.
Daily change monitoring, with a cybersecurity model
Every weekday, the latest frontier models read the day's changes. They then update the attack surface map. A cybersecurity-focused model runs vulnerability review against that day's patches. Shopify names an internal tool here, Dispatch. The post describes Dispatch as an agentic vulnerability scanner. It finds and proves vulnerabilities in the code the team changed. Shopify has not open-sourced Dispatch.
What a reader can take from this post
The 11x figure is a Shopify-internal measurement on a specific rewrite. A team cannot copy the number and expect it in their own repo. The method is the part that travels. Four steps are worth trying this week. Map the attack surface. Write skills per surface. Review at the route level. Run a cybersecurity model on daily diffs. The post links to a second Shopify piece on the harness Dispatch sits inside.
Any team working from mobile OWASP guides already knows the standard. The new part is the cadence: a review that ran once a release now runs once a day.
Source
Building secure mobile apps in the agentic era by Michael Hughes on Shopify Engineering.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.

