Shopify describes Cyber Chaos Monkey, an agentic vulnerability hunter that attacks its own production infrastructure with live credentials. The post covers the Discovery and Sandbox architecture, containment controls such as a credential proxy, egress allowlists, and Kubernetes pod design, and a two-agent judge model that terminates runs that break the Rules of Engagement.
Shopify engineering blog
Loading
Shopify describes a shared generative model trained jointly on search and recommendation data to generate product candidates from semantic-ID tokens. The post covers RQ-KMeans tokenization, training lessons on discovery recall and reinforcement learning, and serving choices such as LFM2-2.6B, constrained decoding, and multi-token prediction, with reported online and offline gains.
Shopify rebuilt its mobile app from React Native to native Swift and Kotlin in 12 weeks with about 30 engineers working alongside AI agents. The post explains the architecture, shared-across-platform patterns, CI guardrails, and review and merge changes that made agent-written code shippable, and reports launch time, memory, and crash-rate results.
Shopify describes Reflex, an MVVM-based architecture that lets AI agents test native iOS and Android apps without a simulator or emulator. Core logic runs headlessly, each ViewModel exposes its state and actions through a JSON introspection contract, and one CLI test suite keeps both platforms in parity. The post explains the module split, the Elm-style ViewModel shape, and the trade-offs against E2E, XCUITest/Espresso, and unit tests.
Shopify describes Remote Dev, a system that runs AI agents against remote macOS machines with iOS simulators and bridged Android emulators, so mobile builds no longer tie up developer laptops. The post covers a warm pool of pre-built Macs, nested-virtualization workarounds for Android on cloud VMs, and a skill-based approach to making agents verify their own changes. It also reports on a classifier loop used to improve agent behavior from failed conversations.
Shopify describes how it post-trains and self-serves open models as task-specific experts for Sidekick's token-heavy subagents. The post walks through a loop covering judge-based quality definition, autoresearch on the frontier agent, model selection, inference benchmarking with AIPerf, and self-healing that turns failed conversations into training data. It reports concrete results, including judge scores rising from 37.7 to 74.5 and roughly 90% lower serving cost than the frontier model.
Shopify describes how it clusters product listings across millions of stores into global Universal Product Identifiers. The post covers a three-stage pipeline of embedding-based candidate retrieval, tiered pair scorers (GBDT, cross-encoder, image-aware reranker), and bounded union-find graph partitioning feeding a learned hierarchical agglomerative merge model tuned for high precision. It also explains how expensive LLM judgments are reused as training data and cached scores.
Shopify describes Overexposed, a scanner that finds Kubernetes Ingresses and Cloud Run services with internet exposure. It explains why reading config files missed what was running and how Cloud Asset Inventory, a careful anonymous probe, and allowlist-aware verdicts give more reliable results. The post also covers SSRF protections, status-code pitfalls, and reducing inconclusive results by about 80 percent.
ShopGym converts live storefronts into anonymized, resettable sandbox shops and generates grounded shopping tasks for agent evaluation. The post explains its exploration, staged generation, verification, and benchmarking workflow, including structural and behavioral comparisons across real and synthetic stores.
Shopify explains how Helix uses LLM agents, incremental checkpoints, behavioral tests, visual comparison, adversarial code reviews, and engineer approval to migrate a 300-screen React Native app to native Swift and Kotlin. The checkpoint-and-gate workflow prioritizes reliable convergence over perfect first attempts while preserving code quality and UI fidelity.
Shopify details how it rebuilt the Shop app from React Native in Swift and Kotlin in 12 weeks with coding agents. The migration improved startup time, Android app size, build times, rendering performance, and session stability while preserving feature parity, analytics, authentication, and notifications.
Shopify explains why improved coding agents changed the trade-offs behind its mobile stack, prompting a move from React Native back to Swift and Kotlin. The post details its greenfield migration strategy, Helix’s checkpoint-based review system, and CLI-driven architecture for faster agent feedback loops.
Shopify explains how River, its Slack-based AI agent, moves vulnerability remediation from detection and patch creation through rebasing, CI validation, human handoff, merge, and verified closure. The workflow revalidates live repository state, binds evidence to current commits, preserves investigation context, and uses deterministic controls for security-critical guarantees.
Based on the title alone, the post presents gisting as a technique for compressing LLM-agent context to increase throughput and reduce inference costs.
We rebuilt our mobile end-to-end testing framework with a strict API and computer vision, raising test stability drastically.
How we compress production failures into model weights every day, beat frontier-model quality, and cut serving costs 96%.
We built an agentic code review and test oracle harness that discovers vulnerabilities, proves them with real tests, and provides Shopify-tuned fixes.
We moved five high-traffic checkout extensions to remote-dom and Polaris web components, cutting bundle sizes drastically and making checkout faster.
How a Hack Days team built a music player, custom GLSL visualizers, and an artist toolkit for storefronts, all in three days.
Building an LLM-powered pipeline to match product listings across millions of merchants into a unified catalog that AI agents can search.