State of CSS/JS 2025 + the problem with spec-driven development

View in browser | Past Issue | Subscribe / Unsubscribe

SitePoint Source

Welcome, Developers! 👋

StateOfJS and WallaceCSS reports show how developers actually use CSS and JavaScript at scale, from feature adoption rates to AI's growing role in codebases. Plus three challenging ideas: CI systems only provide value when failing, spec-driven development that assumes clarity we rarely have, and a future where production code exceeds human readability at scale.

From our sponsor: Superhuman

Find out why 150K+ engineers read The Code twice a week

You're spending 40 hours a week writing code that AI could do in 10.

150k+ engineers at OpenAI, Google & Meta read The Code to learn which AI tools work and how to use them.

Get AI coding techniques from top engineers, workflows that cut your time in half, and tech insights 6 months ahead.

Join 150K+ engineers

🔖 The Reading Room

Articles we have hand-picked for you:

The State of Real-World CSS Usage, 2026 Edition

Project Wallace analyzed 100,000+ websites to reveal how CSS is actually used at scale. The median website ships 309 kB of CSS with 2,802 rules, while modern features like :where (90% adoption) and :has (41% adoption) show strong uptake. However, 18% of sites still use outdated browser hacks, and @layer adoption remains low at just 2.7% despite its usefulness.

By Bart Veneman →

State of JavaScript 2025

TypeScript has emerged as the dominant language with native support across Node.js, Deno, and Bun, while Vite has surpassed webpack in downloads. AI is fundamentally changing development, with nearly 30% of code being AI-generated by late 2025, raising important questions about the future of coding. The focus for 2026 is on faster tools, better developer experience, and technology that augments rather than replaces developer agency.
By Sacha Greif →

The purpose of Continuous Integration is to fail

Continuous integration only provides value when it fails and catches mistakes before deployment. Flaky tests that pass on retry represent an unfixable problem that plagues CI systems. When CI failures no longer reliably indicate real mistakes, the entire value proposition collapses.

By NixCI →

Spec-driven development doesn't work if you're too confused to write the spec

The shift to spec-driven development assumes we know what to specify, but that's rarely true. Writing code is often a discovery process where we learn what we're building by trying to make a computer understand it. Any spec unambiguous enough for an agent to implement is already close enough to code that we could just express it directly. The difficulty isn't typing letters, it's refining our thinking as we formalize fuzzy ideas into precise relationships and rules.

By deontologician →

Write-Only Code

AI is pushing us toward a future where large portions of production code will never be read by humans. Recent improvements in AI mean agents can produce software at volumes that exceed human review capacity. While engineers won't disappear, their role will likely shift from writing and reviewing code to systems design, constraint writing, and risk management.

By Joseph Ruscio →

⏳ Back in Time

Most clicks from last newsletter:

🔗 The Link Lounge

Unordered finds from around the web:

Find something cool? You can send us links to feature here via email.

🧰 The Toolbox

Tools and products we're excited about today:

npmx.dev

npmx.dev is a fast, modern browser for the npm registry offering better developer experience with features like dark mode, code viewing, keyboard navigation, vulnerability warnings, and full URL compatibility with npmjs.com.

Learn more →

Shovel

Shovel is an AI-built open source framework that runs Service Workers as full-stack web applications by implementing browser standards for servers, replacing tools like Express and Vite with a framework-agnostic, standards-based architecture ready for early adopters.

Learn more →

RowBoat

Rowboat is an open-source AI coworker that builds a long-lived knowledge graph from your emails and meetings, then uses that accumulated context to draft documents, prepare meeting briefs, and automate workflows on your local machine.

Learn more →

LiftKit

LiftKit is a UI framework that automatically makes developers better designers by applying golden ratio scaling, optical symmetry, and contrast checks through simple utility classes. Platform-agnostic by design, it currently supports Next.js and is open-source, maintained by Chainlift.

Learn more →

Find out why 150K+ engineers read The Code twice a week

You're spending 40 hours a week writing code that AI could do in 10.

150k+ engineers at OpenAI, Google & Meta read The Code to learn which AI tools work and how to use them.

Get AI coding techniques from top engineers, workflows that cut your time in half, and tech insights 6 months ahead.

Join 150K+ engineers →

🎤 Your Voice

Your feedback shapes what comes next! We read every email, so simply hit reply and tell us what's on your mind.