From Newsgroups to Yahoo!: The Roots of Atomic CSS
Thierry Koblentz, now retired, created Atomic CSS and is widely credited with bringing the approach to mainstream attention through his 2013 Smashing Magazine article, "Challenging CSS Best Practices." Before joining Yahoo!, he taught himself HTML, CSS, and JavaScript as a hobby after moving to the U.S. in 1997, becoming a regular on Macromedia NewsGroups and CSS-Discuss. Front-end stayed a hobby for years — his actual job was as an antique dealer — while he published articles sharing techniques he had learned or discovered.
That work led to a phone call from Yahoo! in 2007 asking him to help build stylesheets for the Yahoo! Site Solutions (YSS) website builder template — a job with no HTML, no JavaScript, just CSS.

CSS Practices Before and After Web Standards
In the early days, Koblentz recalls, there were neither libraries nor published methodologies: "non-semantic" classes, IDs, CSS hacks, conditional comments, frames, CSS expressions, JS sniffing, and designing primarily for Internet Explorer were all common. On his old website he even kept this comment:
<!--MSIE5 Mac needs this comment -->
By the time he joined Yahoo!, the situation had changed. Developers from the U.K. were strong supporters of Web Standards and heavily influenced how markup and CSS were written at the company. Semantic markup was the norm, and CSS followed the Separation of Concern principle strictly — at times, in Koblentz's view, overzealously. YUI had CSS components but no framework yet, and an in-house library called Lego existed though he never used it. OOCSS, SMACSS, ECSS, BEM, Bootstrap, Pure, and others arrived shortly after.
How the Idea Took Shape
Before YSS moved to India, Koblentz's manager Michael Montesano asked whether the incoming Bangalore team could avoid editing the stylesheet, reducing the risk of breakage. The template experience — paying customers complaining about broken pages — had made him wary of any stylesheet change. Koblentz responded with a "utility-sheet" in the spirit of his ez-css library, designed to let developers style pages without adding or editing rules.
A couple of years later, Montesano — by then Director of Engineering — asked whether Yahoo!'s Home Page could be redesigned using utility classes only, given that the team again would not maintain the site. The discussion centered on prioritizing utility classes over semantic classes, which Koblentz believes had not been attempted at that level before. The exercise became a proof of concept, and the resulting static stylesheet, called Stencil, was chosen to redesign Yahoo! My Home Page.

Stencil, Atomizer, and the ACSS Philosophy
Stencil's static nature was good for enforcing a single design style, which the team assumed Yahoo! would adopt company-wide. That did not happen: every design team had its own idea of the perfect font size or margin, and requests for very specific styles kept arriving. The library became unmaintainable, so Koblentz, Renato Iwashima, and Steve Carlson built a tool letting developers create their own styles on the fly while staying atomic. That tool is Atomizer.
With ACSS, the focus shifted from adding CSS declarations to building a rich vocabulary — media queries, descendant selectors, pseudo-classes, and more. Teams could adopt different design styles and style guides with the same library, and developers no longer had to worry about breaking a page or writing selectors to style components. ACSS was built first and foremost to address Yahoo!'s problems in Yahoo!'s environment.
Lessons From Large-Scale CSS
Koblentz learned at Yahoo! how enormous a codebase can be: teams across timezones, hundreds of shared components, third-party code, A/B testing, localization, strict coding standards, and politics. Many techniques that were common practice, he found, were unfit or counter-productive for complex apps. Among the lessons he draws:
- Abstraction can be misleading. A class like
M-10mapped tomargin: 10pxis criticized because changing styling means editing both HTML and CSS — but in large projects that is exactly what happens. If a designer wants a 12px gap and the utility scale only offers 10px or 15px, changingM-3xis impossible without breaking other teams' expectations. Understand the designer's intent: abstraction (e.g. color primary) versus a specific value. If design teams can request any style, avoid naming conventions that lead to ambiguity, since ambiguity leads to breakages. - Styling via document or component structure — combinators like
>or+— assumes markup is immutable, which cannot be assumed in a complex environment. z-indexis even harder when a team does not own the scope of the stack its component lives in.- Qualifying selectors such as
input.requiredversus.input.requiredadd unnecessary specificity (0.1.1 vs. 0.2.0) and block markup changes. - Using the universal selector
*for global styling risks styling someone else's component. - The real problem is not high specificity but the number of specificity levels. Two or three levels are easier to work within than a low-specificity "free for all" that spreads across many levels.
- Do not jump on techniques that haven't been battle tested (remember
will-change?), and know what every declaration triggers —positioncan create a stacking context and a containing block, whileoverflowcan create a block-formatting context.
Koblentz concludes that knowing CSS inside-out is not enough to write it efficiently for a large organization; contradictions with former allies are common. He cites Nicholas Zakas:
Instead of assuming that people are dumb, ignorant, and making mistakes, assume they are smart, doing their best, and that you lack context.
— Nicholas C. Zakas (@slicknet) February 10, 2013
Internal Reception
The My Home Page team accepted ACSS, but momentum elsewhere was slow. A conference call with the Sports team in Santa Monica, aimed at convincing developers that skipping Separation of Concerns would not create chaos, ran into friction; pointing to a piece by Nicolas Gallagher didn't help. The core issue was that the library consisted of utility classes, but its syntax did not ease the conversation. The Mail team did not reject the idea but wanted their own JavaScript approach to use "plain" CSS declarations, unable to stand the ACSS syntax. Data in favor of the library — ~36% less CSS and HTML — spoke for itself, and ACSS was eventually adopted. Seven-plus years later, Yahoo! Home Page, Sports, News, Finance, and other Yahoo! products still use it. Because ACSS makes components page-agnostic, a component copied from Yahoo! Finance into Yahoo! News should look like it belongs.
Syntax and Tooling
For delimiters around property values, the team considered parentheses () and brackets []. Renato recalls that parentheses won for familiarity with JavaScript functions, despite the extra Shift keystroke. The syntax was designed to facilitate automatic rule generation via Atomizer, allow arbitrary or complex styles, and minimize the learning curve.
[<context>[:<pseudo-class>]<combinator>]<Style>[(<value>,<value>?,...)][<!>][:<pseudo-class>][::<pseudo-element>][--<breakpoint_identifier>]
Classes are built on Emmet, so core classes are explicit property/value pairs rather than arbitrary strings. About a dozen helper classes apply multiple declarations — shortcuts such as hiding content from sighted users, or hacks like .Cf for clearfix. A config file lets developers define variables like .PrimaryColor, breakpoints, and more. Detractors call the syntax weird; users call it clever in many ways.
Publishing and Legacy
Having written for many publications before, Koblentz naturally wrote an article about the approach. Yahoo! was sponsoring an October 2013 front-end conference where Renato was scheduled to present the solution, and Koblentz aimed to publish beforehand. He avoided the Yahoo! Developer Network, which lacked comments; A List Apart could not publish in time, but Smashing Magazine accelerated its review. He told Vitaly Friedman that the disclaimer note on the article would sway readers, but did not push back after understanding the reasoning. The comment section drew 200-plus responses, which proved time consuming and frustrating to answer. He finds it amusing that the disclaimer remains now that the methodology is mainstream.
On what he would change, Koblentz notes that pioneers have no others' mistakes to learn from — in French, essuyer les plâtres. The dynamic tooling side offered little inspiration, and it took six years for other libraries to follow suit. One admitted mistake was titling acss.io "Atomic CSS," which, as John Polacek pointed out, created confusion; the title has since changed. His only regret is how the community has treated Atomic CSS/ACSS, including a recent exchange in which someone explained to him what "Atomic CSS" means:
The Atomic CSS library [ACSS] uses the name but I think this is misleading, because the feature you're talking about is the dynamic style generation. "Atomic CSS" as a generic term designates small selectors as atoms, but they're static.
Talk about being erased.



