Claude Code Tips, React Compiler RC, & DoorDash 50% Latency Drop

Hello!

Salesforce's AI writes 20% of the production code, but their developers aren't too worried about it. On the contrary, they’re happy to focus on higher-impact work, while AI handles the basics. That’s just one of the insights from this issue worth your time.

DoorDash shares how a custom search engine helped them lower latency by 50% and cut hardware needs by 75%. The new React Compiler RC automates memoization and brings real improvements to performance. A new guide by Anthropic will show you how to use Claude Code effectively, and the first-ever State of WebDev AI survey will answer any questions you might have about AI in web development.

We also have a practical Vercel cost-cutting guide, and yes, a blog hosted on a Nintendo Wii.

Sounds good? Then relax, grab your coffee, and enjoy reading Frictionless.

In the Queue

• How Pinterest Migrated 3.7 Million Lines of Code

• What Does It Mean to Build in the Modern Enterprise?

• Why OpenAI Dropped Cursor for Windsurf

• Rules for Building Reliable LLM Applications

• Secrets of Event-Hidden Architectures

Reduce Friction

Migrating 3.7 Million Lines of Flow Code to TypeScript

Pinterest’s migration from a stagnating Flow ecosystem wasn’t easy, but moving 3.7 million lines of code to TypeScript saved them from countless upgrade headaches. They pulled it off with more creativity than you’d expect from a company at that scale.

What Makes a Great Developer Experience?

What if developer experience isn’t about fancy tooling, but about protecting attention? Speed alone might not guarantee good work output, so it's time to design work environments that don’t constantly interrupt your team.

How DoorDash’s In-House Search Engine Achieved a 50% Drop in Latency

When Elasticsearch couldn’t keep up with DoorDash’s growing complexity, they built their own search engine on top of Apache Lucene. The result was a 50% lower tail latency, 75% less hardware, and a design worth taking notes on.

Redefining the Fundamentals: What Does It Mean to Build in the Modern Enterprise?

While deep technical knowledge still matters, it’s no longer the most valuable skill in modern enterprises. Many of us are having to relearn how to build, and what “technical” means in a world of high-level tools and AI.

Deepen Your Expertise

Next.js Vercel Cost Optimization Guide

Jakub and I put together this guide for teams looking to cut their Vercel bills. It includes tried-and-true strategies and real code examples we’ve used to help clients save thousands each year. Grab your free copy and start saving.

React Compiler RC

The new stable React Compiler automatically memoizes components, reducing re-renders without useMemo. It adds support for optional chaining, array indexing, SWC builds, and comes with built-in ESLint rules.

JSX Over The Wire

Dan Abramov, co-creator of Redux, explores what it might look like if servers sent JSX instead of JSON, turning APIs into component renderers.

Event-Hidden Architectures

Event-driven systems are powerful and often painful to work with. However, this new approach lets you keep their benefit while hiding the complexity behind a clean interface.

AI Corner

Claude Code: Best Practices For Agentic Coding

The Anthropic team shared tips and tricks for getting the most out of Claude Code, their CLI tool for agentic coding. Learn how to customize environments with CLAUDE.md, manage tool access, and make Claude a part of your workflow.

State of WebDev AI Survey

If you’ve ever wondered how web developers use AI in their workflows, the first-ever State of Web Dev AI survey has answers. It looks into tool adoption, spending habits, what’s working, and where teams are still hitting friction.

This AI Already Writes 20% of Salesforce’s Code. Here’s Why Developers Aren’t Worried

Turns out, coding with AI isn’t the end of the job. At Salesforce, AI writes millions of lines of code, but developers aren’t panicking. Agentforce handles the heavy lifting so engineers can focus on refining and improving the output.

Why OpenAI Wanted to Buy Cursor But Opted for the Fast-Growing Windsurf

Although OpenAI once considered purchasing Cursor, it’s now set its sights on fast-growing rival Windsurf. With a $3B deal reportedly in the works, OpenAI tries to get a stake in the AI coding assistant space.

12 Factor Agents - Principles for Building Reliable LLM Applications

A practical guide to building LLM apps that don’t fall apart in production. Inspired by the 12-Factor App model, it covers things like owning your prompts, managing context windows, and keeping agents focused.

Just Cool

This Blog is Hosted on a Nintendo Wii

Someone turned a 2006 Nintendo Wii into a functioning web server, complete with TLS. It’s ridiculous, brilliant, and somehow... it works.

Let’s Stay in Touch! 📨 

Do you have any comments about this newsletter issue or questions you want to ask? Drop me a message or book a meeting.