Deep dives, system design, and lessons from building commerce infrastructure at scale. By the engineers at devx.
Every internal tool your team ships in 2026 will have an MCP server next to it. This is the production playbook — protocol, code, auth, security, and the mistakes to avoid.
React Native apps ship more bytes than they need to. Here's how we cut bundle size using Hermes, inlineRequires, library replacement, and Android App Bundles — with measurable results per step.
Figma to code has always been the slow part of frontend work. Claude and Figma MCP made it faster, but one extra step made it accurate.
Most of us use AI to write code. But somewhere between the prompt and the output, we lose track of what's actually being built. This is about getting that back.
Cold start in React Native is slow by default. Here's how we cut it from 3.4s to 1.9s on mid-range Android — four targeted changes, no feature rewrites.
We cut LCP by 47% and CLS by 91% on a media-heavy Next.js 16 D2C storefront. Every fix that worked, what we over-engineered, and the priority order that matters.
We built a self-enforcing audit system for React Native that blocks bad code at commit time, auto-generates tests, and learns from every bug it finds.
SPAs moved rendering to the client. Here's what that costs on real mobile hardware, which metrics actually matter, and how to fix the bottleneck.
How one engineer moved 5.6 million files across cloud providers, eliminated 60% duplicate storage, and made every CDN URL unguessable with UUID-based path transformation.
Next.js leaves internal structure entirely up to you. Here's the folder guide I wish existed: from solo project to multi-team monorepo, with reasons.
Most enterprise systems scatter product data, assets, and content across five different tools. Here is how Pimcore collapses that into one governed platform.
AI writes working code in minutes. It also gets the architecture wrong. I share two real enterprise projects where business knowledge caught what AI missed.
Every ecommerce team wastes weeks rebuilding the same admin primitives. Medusa Admin gives you all of them on day one — open source, responsive, yours to own.
I was manually updating prices across 10 Shopify markets every week. One bad night pushed me to build something better.
Our analytics code was duplicated across 15 files, inconsistent across 6 providers, and impossible to debug. Here's the dispatch pattern that lets new devs ship events in 15 minutes.
How we restructured a growing Lambda codebase from flat files into a domain-driven architecture with handler, controller, service layers, Zod validation, and auto-generated scaffolding.
A version mismatch in a shared auth library caused a 3-hour production incident. Here's the monorepo architecture (workspaces, Turborepo task graphs, remote caching) that makes that class of problem impossible.
We replaced UUID with a custom 64-bit time-ordered ID generator in PostgreSQL for our ecommerce platform. Here's the SQL and the trade-offs.
Learn how feature flags let you ship code to production without exposing it to users — enabling safer rollouts, instant rollbacks, and A/B testing.
A friend's team pasted order data into ChatGPT and leaked it. Data sent to AI tools is sensitive. Here is what engineering teams should actually do about it.
AI's accuracy gap in React Native is real and structural. You can't prompt your way past it. Write CLAUDE.md before you write code — that context layer compensates.
A D2C skincare brand outgrew their reviews SaaS. We built a custom Lambda + PostgreSQL replacement in one week: 80% cheaper, sub-100ms responses, fully owned.
Our API stitched data from Shopify and Strapi on every request. During sales, it buckled. Here is how we built a pre-assembled cache that serves full responses in under 100ms.
MicroVMs combine the speed of containers with the strong isolation of VMs, offering lightweight, secure, and efficient compute for modern cloud-native workloads.
Character-based chunking destroyed our policy retrieval accuracy. Section-aware semantic chunking brought it back. Here is exactly how we built it.
Choosing between Expo and Bare React Native? We compare cold start times, bundle sizes, native module flexibility, and developer productivity.
Enterprises lose nearly 25% of every workday to searching for scattered data. We break down how Pimcore brings products, assets, and documents under one roof.
Why the intelligence of modern AI systems comes less from the model itself and more from how we design and structure the context around it.
One giant frontend app. Multiple teams. Constant merge conflicts. Sound familiar? This complete guide to Micro Frontend architecture covers the what, why, how, and when-not-to — with code examples, comparison tables, and an honest take on whether your e-commerce app needs one.
The todo app taught every developer CRUD, REST, and state management. That mental model is now obsolete. Here is what replaced it.
A deep dive into Claude Code's internal architecture - real JSONL session data, 165.9M tokens, multi-model routing, and why 97% of all tokens are cache reads.
I built a 3D diamond ring configurator — 139,968 variants, physically accurate ray bouncing, real chromatic dispersion — using Claude Code, Manus, and ChatGPT as my engineering team.
How a 150-person engineering shop went from AI tool fragmentation to a shared mental model for Claude Code: skills, subagents, and what's next.
Stop looking for the 'best' AI coding assistant. Strategic tool switching saved us 5 weeks on a full-stack 3D platform.
Introducing the devx Engineering Blog - a space where we document the real problems we solve, the systems we build, and the lessons we learn along the way.
