Issue #7 — February 2026

// components/Button.tsx

const Button = ({
  variant = 'ghost',
  children,
  ...props
}) => {
  const base = cn(
    'inline-flex items-center',
    variants[variant]
  );
  return (
    <button className={base}
      {...props}>
      {children}
    </button>
  );
};

← semantic first

Frontend developer mid-sentence during interview, black and white portrait

Marcus Reid, Vercel

"I deleted the entire component library and started from semantic HTML."

— Priya Nair, Issue #4

⬡◈⬟◇⬢◉

Builders
speak.

Compile is a weekly dispatch that sits down with the engineers shaping how the web is built — and asks the questions that don't fit in a README.

start here ↓

Read Issue #1 Free

Joined by 4,200+ engineers this month

The Belief

The best interfaces are built by people, not processes.

Somewhere between the Jira ticket and the pull request, the craft gets lost. Not because engineers stop caring — they don't — but because the conversation around how things are actually made has been replaced by the conversation about how fast they can be shipped.

Compile exists in the gap. It's for the person who reads the changelog and wonders about the decision that didn't make it in. For the tech lead who wants to know how a team of four shipped something that felt like it was built by twenty. For the indie builder who's two hours into a problem and just needs to hear that someone else solved it differently.

"Frontend is the last place where one person's taste can still change everything. We're here to document the people who prove it."

— Compile, Volume 1

Depth over velocity

Every issue takes one person seriously. We don't skim their career — we ask about the one decision that changed how they write code.

Tools are arguments

A framework choice is a philosophy. A linting rule is a value system. We treat the toolchain as a text worth reading closely.

Specificity is kindness

Vague advice helps no one at 2 a.m. We name the exact file, the specific refactor, the real tradeoff. You can use it tomorrow.

The Evidence

Three conversations.
Real decisions.

Issue #1

Marcus Reid

Staff Engineer, Vercel · Brooklyn, NY

"We'd spent eight months building something that was slowing us down."

"The component library we'd spent eight months building was slowing us down more than it was helping. Every new feature required a committee. I made the call on a Friday afternoon — we'd strip it to a set of design tokens and twelve base primitives, and let the product teams own the rest. Monday morning I sent the Slack message. Half the team thought I'd lost my mind."

Component ArchitectureDesign SystemsTeam Structure

22 min readRead full interview →

Issue #3

Priya Nair

Founding Engineer, Linear · San Francisco, CA

"I keep the old repo around as a reminder that complexity has a cost."

"I deleted the entire component library and started from semantic HTML. Not as a statement — I just kept finding myself fighting the abstractions instead of building the product. Three weeks later we had something that loaded in under a second on a 3G connection. The Lighthouse score was almost embarrassing. I keep the old repo around as a reminder that complexity has a cost you don't always see until it's too late."

PerformanceAccessibilitySemantic HTML

31 min readRead full interview →

Issue #6

Tobias Meier

Independent, ex-Figma · Berlin, DE

"The most performant code I've ever written is also the most readable."

"After Figma I took three months off and just read. Not tech blogs — novels, mostly, and a lot of writing about furniture design. There's a Shaker principle: 'Don't make something unless it is both necessary and useful; but if it is both necessary and useful, don't hesitate to make it beautiful.' I come back to that constantly. We've somehow convinced ourselves that 'useful' and 'beautiful' are in tension in software. They're not. They reinforce each other. The most performant code I've ever written is also the most readable. That's not a coincidence."

Craft PhilosophyDesign ThinkingSolo Practice

38 min readRead full interview →

The Workbench

Tools worth arguing about.

Each issue profiles two or three tools that came up in conversation. Not sponsored. Not a roundup. Just the specific things engineers keep mentioning when they talk about how they actually work.

⬡

Biome

Linting & Formatting

v1.6

One tool. Zero config. Replaces ESLint and Prettier with a single Rust-based binary that runs in milliseconds.

↳ Because a linting rule is a value system — and this one respects your time.

◈

Radix UI

Primitives

v1.0

Unstyled, accessible components for building high-quality design systems and web apps.

↳ The primitives argument, made definitive. Bring your own styles, inherit their accessibility work.

⬟

Zod

Schema Validation

v3.22

TypeScript-first schema declaration and validation library with static type inference.

↳ The moment you stop trusting your API responses is the moment Zod pays for itself.

◇

TanStack Query

Server State

Async state management that removes boilerplate and makes data-fetching genuinely pleasant.

↳ Because loading/error/data is a pattern, not a problem — and this solved it.

⬢

Million.js

Performance

v3.1

A drop-in compiler that makes React up to 70% faster by optimizing the virtual DOM.

↳ The most interesting performance bet in the React ecosystem right now.

◉

Playwright

Testing

v1.42

End-to-end testing for modern web apps. Fast, reliable, and works across all browsers.

↳ The test you write at 11pm to prove it still works. This is that test.

Full tool profiles — with usage examples and decision context — appear in each issue.

The Conviction

Craft is not a nostalgic idea.
It's the only way to build things that last.

The engineers who built the interfaces you rely on daily — the ones that feel effortless, that load before you notice them loading, that never make you think — those people made thousands of small decisions that nobody will ever see.

Compile is where those decisions get written down. No hot takes. No growth hacks. Just the quiet, specific work of people who care about how things are made — and are willing to explain why.

Read Issue #1 — right now, free.

Drop your email and we'll send the full inaugural issue instantly. No paywall. No pitch. Just the work.

Weekly. No spam. Unsubscribe whenever. Used by engineers at Vercel, Linear, Shopify, and Stripe.

Compile · Est. 2025 · Built slowly, on purpose

Build­ersspeak.